diff --git a/TensorFlow/contrib/cv/CircleLoss_ID1221_for_TensorFlow/README.md b/TensorFlow/contrib/cv/CircleLoss_ID1221_for_TensorFlow/README.md index 80fab081ca18ca6607186768b01d4584c2330d62..6b71350dbb3b879658dfc7cc6c507aea1caad016 100644 --- a/TensorFlow/contrib/cv/CircleLoss_ID1221_for_TensorFlow/README.md +++ b/TensorFlow/contrib/cv/CircleLoss_ID1221_for_TensorFlow/README.md @@ -1,5 +1,3 @@ -![TensorFlow Requirement: 1.x](https://img.shields.io/badge/TensorFlow%20Requirement-1.x-brightgreen) - # CircleLoss **Circle Loss: A Unified Perspective of Pair Similarity Optimization** @@ -14,57 +12,96 @@ ArXiv: [https://arxiv.org/abs/2002.10857](https://arxiv.org/abs/2002.10857) ![image-20210906114857004](https://picbed-1301760901.cos.ap-guangzhou.myqcloud.com/image-20210906114857004.png) -## 1. 依赖库安装 -见Requirements.txt， 需要安装 numpy， scikit-image， tensorflow等库。 +- 适配昇腾 AI 处理器的实现： + + + https://gitee.com/ascend/ModelZoo-TensorFlow/tree/master/TensorFlow/contrib/cv/CircleLoss_ID1221_for_TensorFlow + + + + -## 2. 数据集 +- 精度 -模型训练采用MARKET1501数据集，将文件夹解压到 /Data/ 下 +| - | 论文 | GPU （v100） | NPU | +|-----|-------|------------|-------| +| acc | 0.997 | 0.941 | 0.942 | -https://www.kaggle.com/pengcw1/market-1501/data +- 性能 -## 3. 训练 +| batchsize |image_size| GPU （v100） | NPU | +|---|---|----|---| +| 30 | 112 x 112| 0.075 s/step | 0.35 s/step | -将模型预训练权重放置到 /checkpoint/ 下 -训练脚本为`train_npu.py`，跳转至项目根目录下，直接执行以下命令即可： +- ## 5.Loss曲线 -```bash -bash script/train.sh -``` +![loss曲线](loss.png) +- 训练超参 + - Batch size: 30 + - Learning rate(LR): 0.01 + - Optimizer: Adam + - Weight decay: 1e-6 + - Loss: Circle loss + +## 支持特性 + +| 特性列表 | 是否支持 | +|-------|------| +| 分布式训练 | 否 | +| 混合精度 | 否 | +| 并行数据 | 否 | + + +## 1. 依赖库安装 -## 4. NPU性能 +见requirements.txt， 需要安装 numpy， scikit-image， tensorflow等库。 +## 2. 模型运行 +- 数据集和预训练模型获取 + 链接：https://pan.baidu.com/s/1Rs3KefgNNELa5V0YNtdlfw + 提取码：vyux + 数据集文件为分卷压缩格式，将所有卷放入同一文件夹解压得到dataset.zip文件 + 将dataset.zip文件解压到代码所在目录即可 + 模型参数文件也保存到当前代码目录 -## 5.Loss曲线 +- 模型精度评估 -![image-20211206125601106](https://picbed-1301760901.cos.ap-guangzhou.myqcloud.com/image-20211206125601106.png) + 1. 配置数据路径参数。 -## 6. 精度对比 + 首先在脚本train_performance_1p.sh中，配置训练数据集路径和预训练模型路径，请用户根据实际路径配置，示例如下： -论文中的测试精度如下所示 + ``` + --lfw_dir ./dataset/lfw-deepfunneled_align + --data_dir ./dataset/MS-Celeb-1M_clean_align + --pretrained_model ./20220602-013518/model-20220602-013518.ckpt-258 + ``` -![image-20211206124710842](https://picbed-1301760901.cos.ap-guangzhou.myqcloud.com/image-20211206124710842.png) + 2. 启动模型精度评估。 -使用NPU训练后得到的测试集精度如下所示： + ``` + bash train_performance_1p.sh + ``` -表格：Market-1501数据集上Circle loss 应用于ResNet50训练的测试结果，其中包括了R-1准确率（%）和mAP（%） -| Method | R-1 | mAP | -| ----------- | ---- | ---- | -| 原论文精度 | 94.2 | 84.9 | -| GPU复现精度 | 81.0 | 62.1 | -| NPU复现精度 | 80.9 | 61.5 | +- 模型性能评估。 -## Contact + 1. 配置数据路径参数。 -Please assign issues to [@louis-yx](https://gitee.com/louis-yx) if you have any question. + 首先在脚本train_1p.sh中，配置数据集路径和模型参数路径，请用户根据实际路径配置，示例如下： -## Credits + ``` + --lfw_dir ./dataset/lfw-deepfunneled_align + --data_dir ./dataset/MS-Celeb-1M_clean_align + --pretrained_model ./20220602-013518/model-20220602-013518.ckpt-258 + ``` -This implementation is derived from [Person_reID_baseline_pytorch](https://github.com/layumi/Person_reID_baseline_pytorch) by [layumi](https://github.com/layumi). + 2. 启动模型性能评估。 + ``` + bash train_1p.sh + ``` diff --git a/TensorFlow/contrib/cv/CircleLoss_ID1221_for_TensorFlow/debug.py b/TensorFlow/contrib/cv/CircleLoss_ID1221_for_TensorFlow/circle_loss.py similarity index 59% rename from TensorFlow/contrib/cv/CircleLoss_ID1221_for_TensorFlow/debug.py rename to TensorFlow/contrib/cv/CircleLoss_ID1221_for_TensorFlow/circle_loss.py index 3735c926a7ead5908ed3fc3f6be5ae2fdaf8f147..2e8d1ffd447c1a581f4eb94db9513c4bc74f41ef 100644 --- a/TensorFlow/contrib/cv/CircleLoss_ID1221_for_TensorFlow/debug.py +++ b/TensorFlow/contrib/cv/CircleLoss_ID1221_for_TensorFlow/circle_loss.py @@ -25,14 +25,33 @@ # WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. # See the License for the specific language governing permissions and # limitations under the License. + +from npu_bridge.npu_init import * import tensorflow as tf -a = tf.constant([1, 2, 3, 4], dtype = tf.float64) -def reducelogsumexp(x): - x = tf.math.exp(x) - x = tf.math.reduce_sum(x) - x = tf.math.log(x) - return x -res=reducelogsumexp(a) -apires=tf.math.reduce_logsumexp(a) -with tf.Session() as sess: - print(sess.run([res,apires])) \ No newline at end of file +import numpy as np +import tensorflow.compat.v1 as tf +tf.disable_eager_execution() + + +def circle_loss(y_true, + y_pred, + gamma = 256, + margin = 0.25, + batch_size=None): + O_p = 1 + margin + O_n = -margin + Delta_p = 1 - margin + Delta_n = margin + if batch_size: + batch_size = batch_size + batch_idxs = tf.expand_dims( + tf.range(0, batch_size, dtype=tf.int32), 1) # shape [batch,1] + alpha_p = tf.nn.relu(O_p - tf.stop_gradient(y_pred)) + alpha_n = tf.nn.relu(tf.stop_gradient(y_pred) - O_n) + # yapf: disable + y_true = tf.cast(y_true, tf.float32) + y_pred = (y_true * (alpha_p * (y_pred - Delta_p)) + + (1 - y_true) * (alpha_n * (y_pred - Delta_n))) * gamma + # yapf: enable + return tf.reduce_mean(tf.nn.softmax_cross_entropy_with_logits(labels=y_true, logits=y_pred)) + diff --git a/TensorFlow/contrib/cv/CircleLoss_ID1221_for_TensorFlow/data.py b/TensorFlow/contrib/cv/CircleLoss_ID1221_for_TensorFlow/data.py deleted file mode 100644 index dcdaedc517312950611f8baa9828c86f1cd498db..0000000000000000000000000000000000000000 --- a/TensorFlow/contrib/cv/CircleLoss_ID1221_for_TensorFlow/data.py +++ /dev/null @@ -1,413 +0,0 @@ -# Copyright 2017 The TensorFlow Authors. All Rights Reserved. -# -# Licensed under the Apache License, Version 2.0 (the "License"); -# you may not use this file except in compliance with the License. -# You may obtain a copy of the License at -# -# http://www.apache.org/licenses/LICENSE-2.0 -# -# Unless required by applicable law or agreed to in writing, software -# distributed under the License is distributed on an "AS IS" BASIS, -# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. -# See the License for the specific language governing permissions and -# limitations under the License. -# ============================================================================ -# Copyright 2021 Huawei Technologies Co., Ltd -# -# Licensed under the Apache License, Version 2.0 (the "License"); -# you may not use this file except in compliance with the License. -# You may obtain a copy of the License at -# -# http://www.apache.org/licenses/LICENSE-2.0 -# -# Unless required by applicable law or agreed to in writing, software -# distributed under the License is distributed on an "AS IS" BASIS, -# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. -# See the License for the specific language governing permissions and -# limitations under the License. - -import os -# from src.dataflow.market import MarketTriplet -# import skimage -# import skimage.transform -import numpy as np - - -def get_file_list(file_dir, file_ext, sub_name=None): - """ Get file list in a directory with sepcific filename and extension - - Args: - file_dir (str): directory of files - file_ext (str): filename extension - sub_name (str): Part of filename. Can be None. - - Return: - List of filenames under ``file_dir`` as well as subdirectories - - """ - re_list = [] - - if sub_name is None: - return np.array([os.path.join(root, name) - for root, dirs, files in os.walk(file_dir) - for name in sorted(files) if name.lower().endswith(file_ext)]) - else: - return np.array([os.path.join(root, name) - for root, dirs, files in os.walk(file_dir) - for name in sorted(files) - if name.lower().endswith(file_ext) and sub_name.lower() in name.lower()]) - - - -# def loadMarketForTest(data_dir='', rescale_im=[384, 128]): -# def normalize_im(im): -# im = skimage.transform.resize( -# im, rescale_im, -# mode='constant', preserve_range=True) -# return np.clip(im/255.0, 0., 1.) - -# query_dir = os.path.join(data_dir, 'query') -# query_data = MarketTripletForTest( -# n_class=None, -# data_dir=query_dir, -# batch_dict_name=['im', 'label','camid'], -# shuffle=True, -# pf=normalize_im) -# # query_data.setup(epoch_val=0, sample_n_class=10, sample_per_class=1) - -# test_dir = os.path.join(data_dir, 'bounding_box_test') -# test_data = MarketTripletForTest( -# n_class=None, -# data_dir=test_dir, -# batch_dict_name=['im', 'label','camid'], -# shuffle=True, -# pf=normalize_im) -# # test_data.setup(epoch_val=0, sample_n_class=750, sample_per_class=2) -# return query_data,test_data - -# query_data,test_data = loadMarketForTest(market_dir, rescale_im=im_size) - - - -# def loadMarket(P,K,data_dir='', rescale_im=[128, 64]): -# def normalize_im(im): -# im = skimage.transform.resize( -# im, rescale_im, -# mode='constant', preserve_range=True) -# return np.clip(im/255.0, 0., 1.) -# -# train_dir = os.path.join(data_dir, 'bounding_box_train') -# train_data = MarketTriplet( -# n_class=None, -# data_dir=train_dir, -# batch_dict_name=['im', 'label'], -# shuffle=True, -# pf=normalize_im) -# train_data.setup(epoch_val=0, sample_n_class=P, sample_per_class=K) -# -# test_dir = os.path.join(data_dir, 'bounding_box_test') -# test_data = MarketTriplet( -# n_class=None, -# data_dir=test_dir, -# batch_dict_name=['im', 'label'], -# shuffle=True, -# pf=normalize_im) -# test_data.setup(epoch_val=0, sample_n_class=P, sample_per_class=K) -# -# data = {} -# data['test'] = test_data -# data['train'] = train_data -# return data - -import numpy as np -import random -import tensorflow as tf - -const_mean = np.zeros([256,128,3]) -const_std = np.zeros([256,128,3]) -mean, std = [0.485, 0.456, 0.406], [0.229, 0.224, 0.225] -for i in range(3): - const_mean[:,:,i] = mean[i] - const_std[:,:,i] = std[i] -const_mean = tf.constant(const_mean,dtype=tf.float32) -const_std = tf.constant(const_std,dtype=tf.float32) - -# mean_tf1 = mean[0]*tf.ones(shape=[256,128,1]) -# mean_tf2 = mean[1]*tf.ones(shape=[256,128,1]) -# mean_tf3 = mean[2]*tf.ones(shape=[256,128,1]) -# std_tf1 = std[0]*tf.ones(shape=[256,128,1]) -# std_tf2 = std[1]*tf.ones(shape=[256,128,1]) -# std_tf3 = std[2]*tf.ones(shape=[256,128,1]) -# mean = tf.concat([mean_tf1,mean_tf2,mean_tf3],axis=2) -# std = tf.concat([std_tf1,std_tf2,std_tf3],axis=2) - -def norm_trans(im): - im = (im-const_mean)/const_std - return im -def random_erasing(img, probability = 0.5, sl = 0.02, sh = 0.4, r1 = 0.3): - ''' - img is a 3-D variable (ex: tf.Variable(image, validate_shape=False) ) and HWC order - ''' - # HWC order - import random - if random.uniform(0, 1) > probability: - return img - # print('erasing') - height = int(img.shape[0]) - width = int(img.shape[1]) - channel = int(img.shape[2]) - area = height*width - - erase_area_low_bound = round(np.sqrt(sl * area * r1)) - erase_area_up_bound = round((sh * area) / r1) - h_upper_bound = min(erase_area_up_bound, height) - w_upper_bound = min(erase_area_up_bound, width) - - h = np.random.randint(low=erase_area_low_bound,high=h_upper_bound) - w = np.random.randint(low=erase_area_low_bound,high=w_upper_bound) - - x1 = np.random.randint(low=0,high=height+1-h) - y1 = np.random.randint(low=0,high=width+1-w) - - mask = np.ones([height,width,channel]) - mask[x1:x1+h, y1:y1+w, :] = 0 - # print(f'erasing {h}x{w} pixel!') - masktensor = tf.constant(mask,dtype=tf.float32) - return img*masktensor - - - - -# from src.utils.dataflow import get_file_list -def tf_dataset_PK(P=4,K=4,erasing_prob=0.5,resize_image=[256,128],epoch_size=60,path='/home/nanshen/xutan/yixin/market1501/Market-1501-v15.09.15/bounding_box_train'): - imlist = get_file_list(path, '.jpg') - print(len(imlist)) - label = [] - for im_path in imlist: - import ntpath - head, tail = ntpath.split(im_path) - class_label = tail.split('_')[0] - label.append(int(class_label)) - from sklearn import preprocessing - le = preprocessing.LabelEncoder() - le.fit(label) - label_ = le.transform(label) - # print(f'加载数据中一共有{len(le.classes_)}个类') - # 函数的功能时将filename对应的图片文件读进来，并缩放到统一的大小 - def _parse_function(filename, label): - image_string = tf.read_file(filename) - image_decoded = tf.image.decode_jpeg(image_string,channels=3) - image_resized = tf.image.resize(image_decoded, resize_image,method=tf.image.ResizeMethod.BICUBIC) - # print(image_resized.shape) - # im = np.pad(image_resized,pad_width=((10,10),(10,10),(0,0)),mode='constant') - image_padded = tf.image.pad_to_bounding_box(image_resized, 10,10,resize_image[0]+20,resize_image[1]+20) - # im = random_crop(im,[256, 128]) - crop = tf.image.random_crop(image_padded,[resize_image[0],resize_image[1],3]) - rd = np.random.rand() - if rd < 0.5: - flip = tf.image.flip_left_right(crop) - else: - flip = crop - im = tf.clip_by_value(flip/255.0,0,1.0) - # norm - im = norm_trans(im) - # erasing - im = random_erasing(im,probability=erasing_prob) - return im, label - - imlists_dict = {} - for i in range(751): - imlists_dict[i] = [] - for indxi,labeli in enumerate(label_): - imlists_dict[labeli].append(imlist[indxi]) - - datasets = [] - for i in range(751): - imlist_ind = imlists_dict[i] - # 图片文件的列表 - filenames = tf.constant(imlist_ind) - # label[i]就是图片filenames[i]的label - label_ind = [i]*len(imlist_ind) - labels = tf.constant(label_ind) - - # 此时dataset中的一个元素是(filename, label) - dataset = tf.data.Dataset.from_tensor_slices((filenames, labels)) - - # 此时dataset中的一个元素是(image_resized, label) - dataset = dataset.map(_parse_function) - - # 此时dataset中的一个元素是(image_resized_batch, label_batch) - dataset = dataset.shuffle(buffer_size=1000).batch(K).repeat(epoch_size) - - datasets.append(dataset) - - return [len(imlist),datasets,le] - -# from src.utils.dataflow import get_file_list -def tf_dataset(erasing_prob=0.5,resize_image=[256,128],epoch_size=60,batch_size=32,path='/home/nanshen/xutan/yixin/market1501/Market-1501-v15.09.15/bounding_box_train'): - imlist = get_file_list(path, '.jpg') - print(len(imlist)) - label = [] - for im_path in imlist: - import ntpath - head, tail = ntpath.split(im_path) - class_label = tail.split('_')[0] - label.append(int(class_label)) - from sklearn import preprocessing - le = preprocessing.LabelEncoder() - le.fit(label) - label_ = le.transform(label) - # print(f'加载数据中一共有{len(le.classes_)}个类') - # 函数的功能时将filename对应的图片文件读进来，并缩放到统一的大小 - def _parse_function(filename, label): - image_string = tf.read_file(filename) - image_decoded = tf.image.decode_jpeg(image_string,channels=3) - image_resized = tf.image.resize(image_decoded, resize_image,method=tf.image.ResizeMethod.BICUBIC) - # print(image_resized.shape) - # im = np.pad(image_resized,pad_width=((10,10),(10,10),(0,0)),mode='constant') - image_padded = tf.image.pad_to_bounding_box(image_resized, 10,10,resize_image[0]+20,resize_image[1]+20) - # im = random_crop(im,[256, 128]) - crop = tf.image.random_crop(image_padded,[resize_image[0],resize_image[1],3]) - rd = np.random.rand() - if rd < 0.5: - flip = tf.image.flip_left_right(crop) - else: - flip = crop - im = tf.clip_by_value(flip/255.0,0,1.0) - # norm - im = norm_trans(im) - # erasing - im = random_erasing(im,probability=erasing_prob) - return im, label - - # 图片文件的列表 - filenames = tf.constant(imlist) - # label[i]就是图片filenames[i]的label - labels = tf.constant(label_) - - # 此时dataset中的一个元素是(filename, label) - dataset = tf.data.Dataset.from_tensor_slices((filenames, labels)) - - # 此时dataset中的一个元素是(image_resized, label) - dataset = dataset.map(_parse_function) - - # 此时dataset中的一个元素是(image_resized_batch, label_batch) - dataset = dataset.shuffle(buffer_size=1000).batch(batch_size).repeat(epoch_size) - - return [len(imlist),dataset,le] - -# from src.utils.dataflow import get_file_list -def tf_val_dataset(resize_image=[256,128],epoch_size=60,batch_size=32,datapath='/home/liulizhao/projects/liuyixin/Market-1501-v15.09.15/'): - queryimlist = get_file_list(datapath+'query', '.jpg') - galleyimlist = get_file_list(datapath+'bounding_box_test', '.jpg') - imlist = list(queryimlist) + list(galleyimlist) - label = [] - for im_path in imlist: - import ntpath - head, tail = ntpath.split(im_path) - class_label = tail.split('_')[0] - label.append(int(class_label)) - from sklearn import preprocessing - le = preprocessing.LabelEncoder() - le.fit(label) - label_ = le.transform(label) - # 函数的功能时将filename对应的图片文件读进来，并缩放到统一的大小 - def _parse_function(filename, label): - image_string = tf.read_file(filename) - image_decoded = tf.image.decode_jpeg(image_string,channels=3) - image_resized = tf.image.resize(image_decoded, resize_image,method=tf.image.ResizeMethod.BICUBIC) - # print(image_resized.shape) - # im = np.pad(image_resized,pad_width=((10,10),(10,10),(0,0)),mode='constant') - # image_padded = tf.image.pad_to_bounding_box(image_resized, 10,10,resize_image[0]+20,resize_image[1]+20) - # im = random_crop(im,[256, 128]) - # crop = tf.image.random_crop(image_padded,[resize_image[0],resize_image[1],3]) - # rd = np.random.rand() - # if rd < 0.5: - # flip = tf.image.flip_left_right(crop) - # else: - # flip = crop - im = tf.clip_by_value(image_resized/255.0,0,1.0) - # norm - im = norm_trans(im) - return im, label - - # 图片文件的列表 - filenames = tf.constant(imlist) - # label[i]就是图片filenames[i]的label - labels = tf.constant(label_) - - # 此时dataset中的一个元素是(filename, label) - dataset = tf.data.Dataset.from_tensor_slices((filenames, labels)) - - # 此时dataset中的一个元素是(image_resized, label) - dataset = dataset.map(_parse_function) - - # 此时dataset中的一个元素是(image_resized_batch, label_batch) - dataset = dataset.shuffle(buffer_size=1000).batch(batch_size).repeat(epoch_size) - - return dataset - - -import random -def random_crop(image, crop_shape): - img_h = image.shape[0] - img_w = image.shape[1] - # img_d = image.shape[2] - nh = random.randint(0, img_h - crop_shape[0]) - nw = random.randint(0, img_w - crop_shape[1]) - image_crop = image[nh:nh + crop_shape[0], nw:nw + crop_shape[1],:] - return image_crop - -# def loadMarketForTrain(P,K,data_dir='', rescale_im=[128, 64]): -# def process_im(im): -# im = skimage.transform.resize( -# im, rescale_im, -# mode='constant', preserve_range=True) -# # pad -# im = np.pad(im,pad_width=((10,10),(10,10),(0,0)),mode='constant') -# # random crop -# im = random_crop(im,rescale_im) -# # horizontalFlip -# im = np.fliplr(im) -# # to tensor -# im = np.clip(im/255.0, 0., 1.) -# # normalize -# mean, std = [0.485, 0.456, 0.406], [0.229, 0.224, 0.225] -# for i in range(3): -# im[:,:,i] = (im[:,:,i]-mean[i])/std[i] -# return im -# -# train_dir = os.path.join(data_dir, 'bounding_box_train') -# train_data = MarketTriplet( -# n_class=None, -# data_dir=train_dir, -# batch_dict_name=['im', 'label'], -# shuffle=True, -# pf=process_im) -# train_data.setup(epoch_val=0, sample_n_class=P, sample_per_class=K) -# -# return train_data - -def feed_dict(data,train,x,y_,lr,lr_new): - """定义训练和测试操作""" - if train: - dict_ = data['train'].next_batch_dict() - xs = dict_['im'] - ys = dict_['label'] - else: - dict_ = data['test'].next_batch_dict() - xs = dict_['im'] - ys = dict_['label'] - return {x: xs, y_: ys,lr:lr_new} - -def get_next_batch(data): - dict_ = data.next_batch_dict() - xs = dict_['im'] - ys = dict_['label'] - return xs,ys - -if __name__ == '__main__': - valset = tf_val_dataset(batch_size=128,epoch_size=None) - testiterator = valset.make_one_shot_iterator() - testnext = testiterator.get_next() - \ No newline at end of file diff --git a/TensorFlow/contrib/cv/CircleLoss_ID1221_for_TensorFlow/data/learning_rate_circle_loss.txt b/TensorFlow/contrib/cv/CircleLoss_ID1221_for_TensorFlow/data/learning_rate_circle_loss.txt new file mode 100644 index 0000000000000000000000000000000000000000..87d6bfbb41695c2d89b50d9864f84c63e80fa112 --- /dev/null +++ b/TensorFlow/contrib/cv/CircleLoss_ID1221_for_TensorFlow/data/learning_rate_circle_loss.txt @@ -0,0 +1,6 @@ +# Learning rate schedule +# Maps an epoch number to a learning rate +0: 0.1 +400: 0.01 +700: 0.001 +900: 0.0001 \ No newline at end of file diff --git a/TensorFlow/contrib/cv/CircleLoss_ID1221_for_TensorFlow/data/pairs.txt b/TensorFlow/contrib/cv/CircleLoss_ID1221_for_TensorFlow/data/pairs.txt new file mode 100644 index 0000000000000000000000000000000000000000..0b3cb5482054061da4c90e8985b6a7fd18e437cf --- /dev/null +++ b/TensorFlow/contrib/cv/CircleLoss_ID1221_for_TensorFlow/data/pairs.txt @@ -0,0 +1,6001 @@ +10 300 +Abel_Pacheco 1 4 +Akhmed_Zakayev 1 3 +Akhmed_Zakayev 2 3 +Amber_Tamblyn 1 2 +Anders_Fogh_Rasmussen 1 3 +Anders_Fogh_Rasmussen 1 4 +Angela_Bassett 1 5 +Angela_Bassett 2 5 +Angela_Bassett 3 4 +Ann_Veneman 3 5 +Ann_Veneman 6 10 +Ann_Veneman 10 11 +Anthony_Fauci 1 2 +Antony_Leung 1 2 +Antony_Leung 2 3 +Anwar_Ibrahim 1 2 +Augusto_Pinochet 1 2 +Barbara_Brezigar 1 2 +Benjamin_Netanyahu 1 4 +Benjamin_Netanyahu 4 5 +Bernard_Law 2 3 +Bernard_Law 3 4 +Bertrand_Bonello 1 2 +Bill_Frist 2 9 +Bill_Frist 4 5 +Bob_Graham 2 4 +Bob_Graham 3 6 +Bob_Graham 4 6 +Bob_Graham 5 6 +Boris_Becker 2 6 +Brad_Johnson 1 3 +Brad_Johnson 2 3 +Brad_Johnson 2 4 +Brian_Griese 1 2 +Candice_Bergen 1 2 +Candice_Bergen 2 3 +Carla_Myers 1 2 +Cathy_Freeman 1 2 +Chang_Dae-whan 1 2 +Charles_Grassley 1 2 +Clare_Short 1 3 +Clare_Short 1 4 +Corinne_Coman 1 2 +Dai_Bachtiar 1 2 +Dale_Earnhardt_Jr 1 3 +David_Caruso 1 2 +Demi_Moore 1 3 +Dick_Vermeil 1 2 +Doris_Roberts 1 2 +Doris_Roberts 1 3 +Drew_Barrymore 1 2 +Edmund_Stoiber 4 7 +Edmund_Stoiber 5 7 +Edmund_Stoiber 7 12 +Edmund_Stoiber 9 11 +Elinor_Caplan 1 2 +Emmanuelle_Beart 1 2 +Emmanuelle_Beart 1 3 +Federico_Trillo 1 2 +Federico_Trillo 1 3 +Federico_Trillo 2 3 +Francis_Ford_Coppola 1 2 +Fred_Thompson 1 2 +Fred_Thompson 1 3 +Fred_Thompson 2 3 +Garry_Trudeau 1 2 +Gary_Williams 1 2 +Gene_Robinson 1 2 +Gene_Robinson 1 3 +Gene_Robinson 2 3 +George_Galloway 1 3 +George_Galloway 2 3 +George_Galloway 2 4 +Geraldine_Chaplin 1 2 +Geraldine_Chaplin 1 4 +Geraldine_Chaplin 3 4 +Gerardo_Gambala 1 2 +Gian_Marco 1 2 +Gian_Marco 2 3 +Gillian_Anderson 1 2 +Gordon_Brown 6 9 +Gordon_Brown 10 13 +Grant_Hackett 1 2 +Grant_Hackett 1 3 +Grant_Hackett 1 5 +Gray_Davis 7 22 +Gray_Davis 21 26 +Gregg_Popovich 1 5 +Gregg_Popovich 3 4 +Guy_Ritchie 1 2 +Hamzah_Haz 1 2 +Harbhajan_Singh 1 2 +Hootie_Johnson 1 2 +Hugo_Chavez 2 27 +Hugo_Chavez 27 53 +Hugo_Chavez 36 51 +Hugo_Chavez 41 50 +Hugo_Chavez 45 56 +Isaiah_Washington 1 2 +Jack_Grubman 1 2 +Jacques_Rogge 1 9 +Jacques_Rogge 3 7 +Jacques_Rogge 3 10 +Jacques_Rogge 4 6 +Jacques_Rogge 4 9 +Jacques_Rogge 5 8 +Jacques_Rogge 6 8 +James_Caan 1 2 +James_Caan 1 3 +James_Caan 2 3 +James_Jones 1 2 +Jamling_Norgay 1 2 +Janica_Kostelic 1 2 +Jeanne_Moreau 1 2 +Jeffrey_Archer 1 2 +Jennifer_Garner 1 3 +Jennifer_Garner 4 9 +Jennifer_Garner 5 6 +Jennifer_Garner 5 9 +Jennifer_Garner 7 12 +Jennifer_Garner 8 11 +Jeremy_Greenstock 2 24 +Jeremy_Greenstock 3 22 +Jeremy_Greenstock 5 11 +Jeremy_Greenstock 10 14 +Jeremy_Greenstock 17 21 +Jessica_Lange 1 2 +Jimmy_Carter 1 7 +Jimmy_Carter 4 6 +Jimmy_Carter 5 8 +Jimmy_Carter 6 7 +Jodie_Foster 1 2 +John_Manley 1 6 +John_Manley 2 5 +John_McCain 2 3 +John_McCain 3 7 +John_McCain 6 7 +John_Snow 3 15 +John_Snow 4 16 +John_Snow 9 11 +Johnson_Panjaitan 1 2 +Jorge_Castaneda 1 2 +Jose_Serra 1 9 +Jose_Serra 2 8 +Jose_Serra 3 5 +Jose_Serra 3 7 +Jose_Serra 6 9 +Jose_Theodore 1 2 +Joseph_Blatter 1 2 +Joseph_Ralston 1 2 +Juan_Pablo_Montoya 1 8 +Juan_Pablo_Montoya 2 4 +Juan_Pablo_Montoya 4 5 +Julianna_Margulies 1 2 +Justin_Guarini 1 2 +Justin_Guarini 2 3 +Justine_Henin 2 3 +Kate_Winslet 1 2 +Kate_Winslet 1 4 +Kate_Winslet 2 3 +Kate_Winslet 2 4 +Kathy_Winters 1 2 +Kelvin_Sampson 1 2 +Kevin_Stallings 1 2 +Kim_Jong-Il 2 3 +Kristin_Davis 1 2 +Kristin_Davis 2 3 +Lance_Armstrong 1 17 +Lance_Armstrong 5 10 +Lance_Armstrong 7 11 +Lance_Armstrong 10 18 +Lance_Armstrong 14 17 +Laurent_Jalabert 1 2 +Lindsey_Graham 1 2 +Lisa_Gottsegen 1 2 +Lleyton_Hewitt 1 13 +Lleyton_Hewitt 16 27 +Lleyton_Hewitt 18 37 +Lleyton_Hewitt 24 36 +Luis_Ernesto_Derbez_Bautista 3 6 +Mario_Dumont 1 2 +Mark_Wahlberg 1 3 +Mark_Wahlberg 2 3 +Mark_Wahlberg 2 4 +Marlene_Weingartner 1 2 +Martha_Bowen 1 2 +Martin_Cauchon 1 2 +Martin_Hoellwarth 1 2 +Martin_Sheen 1 2 +Matthew_Broderick 1 4 +Matthew_Broderick 2 3 +Mike_Krzyzewski 1 6 +Mikhail_Gorbachev 1 2 +Monica_Bellucci 1 4 +Monica_Bellucci 2 4 +Naomi_Campbell 1 2 +Nestor_Kirchner 3 30 +Nestor_Kirchner 12 21 +Nestor_Kirchner 17 25 +Nicolas_Lapentti 1 2 +Noah_Wyle 2 3 +Nora_Bendijo 1 2 +Nursultan_Nazarbayev 1 2 +Pamela_Anderson 1 2 +Pamela_Anderson 1 5 +Patricia_Heaton 1 2 +Patty_Schnyder 1 2 +Patty_Schnyder 2 3 +Patty_Schnyder 2 4 +Patty_Schnyder 3 4 +Paul_Bremer 6 17 +Paul_Bremer 7 10 +Paul_Bremer 9 15 +Paul_Bremer 14 16 +Paul_Sarbanes 1 3 +Paul_Sarbanes 2 3 +Paul_William_Hurley 1 2 +Pervez_Musharraf 1 6 +Pervez_Musharraf 2 16 +Pervez_Musharraf 3 10 +Pervez_Musharraf 4 14 +Pervez_Musharraf 5 7 +Pervez_Musharraf 5 15 +Phil_Gramm 1 2 +Prince_Edward 1 2 +Ricardo_Lagos 3 12 +Ricardo_Lagos 23 27 +Ricardo_Monasterio 1 2 +Rich_Gannon 1 2 +Rick_Perry 1 4 +Rick_Perry 1 5 +Rick_Perry 2 4 +Rick_Perry 3 4 +Robert_Bonner 2 3 +Robert_Evans 2 3 +Robert_Fico 1 2 +Romano_Prodi 3 5 +Roy_Moore 2 5 +Roy_Moore 3 6 +Sachiko_Yamada 1 2 +Sachiko_Yamada 1 4 +Sachiko_Yamada 2 3 +Sachiko_Yamada 2 4 +Saeb_Erekat 1 2 +Sam_Bith 1 2 +Sam_Bith 1 3 +Sam_Bith 2 3 +Sean_Hayes 1 2 +Sergio_Garcia 1 2 +Silvia_Farina_Elia 1 2 +Steve_Ballmer 1 2 +Steve_Ballmer 1 3 +Stockard_Channing 1 3 +Stockard_Channing 2 3 +Terry_McAuliffe 1 2 +Terry_McAuliffe 1 3 +Thomas_Rupprath 1 3 +Thomas_Rupprath 2 3 +Tom_Ridge 6 16 +Tom_Ridge 8 25 +Tom_Ridge 9 26 +Tom_Ridge 18 22 +Tommy_Haas 4 5 +Tommy_Thompson 1 2 +Tommy_Thompson 1 7 +Tommy_Thompson 2 8 +Tommy_Thompson 4 8 +Tommy_Thompson 6 9 +Tomoko_Hagiwara 1 2 +Trent_Lott 1 2 +Trent_Lott 2 3 +Trent_Lott 2 8 +Trent_Lott 6 11 +Trent_Lott 7 10 +Trent_Lott 8 16 +Tsutomu_Takebe 1 2 +Tyler_Hamilton 1 2 +Tyra_Banks 1 2 +Vaclav_Havel 2 4 +Vaclav_Havel 2 5 +Vaclav_Havel 2 6 +Vaclav_Havel 4 9 +Vaclav_Havel 5 7 +Valerie_Harper 1 2 +Vince_Carter 3 4 +Vincent_Brooks 2 3 +Vincent_Brooks 2 7 +Vincent_Brooks 4 5 +Vincent_Brooks 4 7 +Vincent_Gallo 1 2 +Vitali_Klitschko 1 2 +Vitali_Klitschko 3 4 +William_Macy 1 4 +William_Macy 2 4 +William_Macy 3 4 +Woody_Allen 2 4 +Woody_Allen 3 5 +Yukiko_Okudo 1 2 +Zico 1 2 +Zico 2 3 +Abdel_Madi_Shabneh 1 Dean_Barker 1 +Abdel_Madi_Shabneh 1 Giancarlo_Fisichella 1 +Abdel_Madi_Shabneh 1 Mikhail_Gorbachev 1 +Abdul_Rahman 1 Portia_de_Rossi 1 +Abel_Pacheco 1 Jong_Thae_Hwa 2 +Abel_Pacheco 2 Jean-Francois_Lemounier 1 +Afton_Smith 1 Dwayne_Wade 1 +Ahmad_Jbarah 1 James_Comey 1 +Akhmed_Zakayev 2 Donna_Morrissey 1 +Alan_Dershowitz 1 Bertrand_Bonello 1 +Alanis_Morissette 1 Martin_Cauchon 1 +Alexander_Lukashenko 1 Heather_Chinnock 1 +Alfonso_Cuaron 1 Jason_Priestley 1 +Alfonso_Cuaron 1 Patty_Schnyder 2 +Alfonso_Soriano 1 Bill_Nelson 2 +Alfonso_Soriano 1 Julio_De_Brun 1 +Alfonso_Soriano 1 Patty_Schnyder 3 +Alonzo_Mourning 1 Cecilia_Cheung 1 +Amber_Tamblyn 2 Benjamin_Netanyahu 1 +Amporn_Falise 1 Joe_Pantoliano 1 +Anders_Fogh_Rasmussen 2 Johnson_Panjaitan 2 +Andre_Bucher 1 Joseph_Ralston 1 +Andre_Bucher 1 Maria_Garcia 1 +Andrew_Gilligan 1 Henry_Castellanos 1 +Andrew_Shutley 1 Edmund_Stoiber 5 +Andrew_Shutley 1 Mitchell_Swartz 1 +Andrew_Shutley 1 Saeb_Erekat 2 +Andy_Dick 1 Simon_Yam 1 +Andy_Griffith 1 Osrat_Iosef 1 +Andy_Wisecarver 1 Dimitar_Berbatov 1 +Angela_Lansbury 1 Steven_Van_Zandt 1 +Angela_Lansbury 2 John_Coomber 1 +Ann_Veneman 6 Sergio_Garcia 2 +Ann_Veneman 8 Ted_Williams 1 +Annie-Jeanne_Reynaud 1 SJ_Twu 1 +Anthony_Carter 1 Eliza_Dushku 1 +Antonio_Cassano 1 Paul_Celluci 1 +Anwar_Ibrahim 1 David_Alpay 1 +Armand_Sargen 1 Daryl_Sabara 1 +Armand_Sargen 1 Kelvin_Sampson 3 +Armand_Sargen 1 Lisa_Gottsegen 2 +Atom_Egoyan 1 Bill_Stapleton 1 +Atom_Egoyan 1 Janis_Ruth_Coulter 1 +Barbara_Brezigar 2 Doris_Roberts 2 +Barbara_Felt-Miller 1 Leticia_Dolera 1 +Bart_Freundlich 1 Ernie_Grunfeld 1 +Bart_Freundlich 1 Kirsten_Gilham 1 +Benjamin_Martinez 1 Garry_McCoy 1 +Benjamin_Netanyahu 1 Maria_Callas 1 +Benjamin_Netanyahu 5 Frank_Beamer 1 +Bernard_Law 1 Liu_Xiaoqing 1 +Bernard_Law 3 Valerie_Harper 2 +Bertrand_Bonello 1 Jong_Thae_Hwa 2 +Bill_Bradley 1 Chen_Tsai-chin 1 +Bill_Bradley 1 Helen_Alvare 1 +Bill_Elliott 1 Mary_Anne_Souza 1 +Bill_Frist 5 Jimmy_Kimmel 2 +Bill_Maher 1 Brad_Russ 1 +Bill_Maher 1 Juliette_Binoche 1 +Bill_Nelson 1 Kim_Jong-Il 3 +Bill_Nelson 2 Gillian_Anderson 1 +Bill_Stapleton 1 Sedigh_Barmak 1 +Bill_Stapleton 1 Valerie_Harper 1 +Billy_Bob_Thornton 1 Herb_Dhaliwal 1 +Billy_Bob_Thornton 1 Nong_Duc_Manh 1 +Bob_Alper 1 Kevin_Millwood 1 +Bob_Graham 2 Dwayne_Wade 1 +Bob_Petrino 1 Geraldine_Chaplin 4 +Bob_Petrino 1 Jorge_Castaneda 1 +Boris_Becker 4 Julianna_Margulies 2 +Brad_Russ 1 Hana_Urushima 1 +Brad_Russ 1 Romeo_Gigli 1 +Brawley_King 1 Tom_Glavine 2 +Brian_Griese 2 Jeffrey_Archer 2 +Brian_Griese 2 Laura_Elena_Harring 1 +Brian_Griese 2 Nicolas_Lapentti 2 +Bryan_Adams 1 Michael_Kors 1 +Bryan_Adams 1 Mohamed_Seineldin 1 +Calbert_Cheaney 1 Ian_Smith 1 +Calbert_Cheaney 1 Robert_Downey_Jr 1 +Carl_Reiner 1 Hamid_Efendi 1 +Carl_Reiner 2 John_Engler 1 +Carl_Reiner 2 Prince_Rainier_III 1 +Carl_Reiner 2 Tom_Glavine 2 +Carlo_Azeglio_Ciampi 1 Francis_Ford_Coppola 1 +Carlos_Arroyo 1 Shane_Phillips 1 +Carlos_Paternina 1 Emily_Stevens 1 +Carlos_Paternina 1 Paul_Sarbanes 1 +Casey_Mears 1 Mike_Davis 1 +Casey_Mears 1 Yukiko_Okudo 1 +Cathy_Freeman 2 William_Martin 2 +Cecilia_Cheung 1 Daryl_Parks 1 +Cecilia_Cheung 1 Pascal_Affi_Nguessan 1 +Chen_Tsai-chin 1 Dereck_Whittenburg 1 +Chen_Tsai-chin 1 Mamdouh_Habib 1 +Cho_Myung-kyun 1 David_Bell 1 +Cho_Myung-kyun 1 Fernando_Sanz 1 +Cho_Myung-kyun 1 Georgia_Giddings 1 +Cho_Myung-kyun 1 Richard_Fine 1 +Choi_Yun-yong 1 Chuck_Eidson 1 +Chris_Dodd 1 Taylor_Twellman 1 +Chris_Swecker 1 Tom_Vilsack 1 +Christian_Lacroix 1 Laura_Elena_Harring 1 +Christian_Lacroix 1 Ornella_Muti 1 +Chuck_Eidson 1 Sigourney_Weaver 1 +Clare_Short 4 Don_Carcieri 1 +Coco_dEste 1 Darvis_Patton 1 +Coco_dEste 1 Melina_Kanakaredes 1 +Coco_dEste 1 Tom_Rouen 1 +Coleen_Rowley 1 Nong_Duc_Manh 1 +Corinne_Coman 2 Frank_Beamer 1 +Dale_Earnhardt_Jr 1 Nick_Reilly 1 +Dario_Franchitti 1 Henry_Castellanos 1 +Darren_Campel 1 Hilary_McKay 1 +Darvis_Patton 1 Gerard_Tronche 1 +Darvis_Patton 1 William_Macy 4 +Daryl_Parks 1 Guus_Hiddink 1 +Daryl_Sabara 1 Nick_Reilly 1 +Daryl_Sabara 1 Valentina_Tereshkova 1 +Dave_Johnson 1 Howard_Stern 1 +Dave_Tucker 1 Gary_Gitnick 1 +David_Collenette 1 Salman_Khan 1 +David_Westerfield 1 Stan_Kroenke 1 +Dean_Jacek 1 Larry_Wilmore 1 +Demi_Moore 2 Fred_Thompson 1 +Demi_Moore 2 Linus_Roache 1 +Dereck_Whittenburg 1 Lindsey_Graham 2 +Dianne_Reeves 1 Larry_Wilmore 1 +Dianne_Reeves 1 Romeo_Gigli 1 +Don_Carcieri 1 Janica_Kostelic 1 +Dora_Bakoyianni 1 Richard_Sambrook 2 +Dora_Bakoyianni 1 Saeb_Erekat 2 +Doris_Roberts 1 Nong_Duc_Manh 1 +Doug_Wilson 1 Szu_Yu_Chen 1 +Douglas_Gansler 1 Martin_Brooke 1 +Douglas_Gansler 1 Ronald_Kadish 1 +Dwayne_Wade 1 Mike_Farrar 1 +Edward_Arsenault 1 Jim_Hardin 1 +Einars_Repse 1 Minnie_Mendoza 1 +Einars_Repse 1 Tim_Blake_Nelson 1 +Elinor_Caplan 1 Hilary_McKay 1 +Eliza_Dushku 1 George_Lucas 1 +Eliza_Dushku 1 Itzhak_Perlman 1 +Emily_Stevens 1 Janez_Drnovsek 1 +Emmanuelle_Beart 2 Phil_Jackson 1 +Eric_Daze 1 Sterling_Hitchcock 1 +Erika_Christensen 2 Michael_Dell 1 +Erika_Christensen 2 Woody_Allen 2 +Eriko_Tamura 1 Georgia_Giddings 1 +Ernie_Grunfeld 1 Frank_Coraci 1 +Eugene_Melnyk 1 Mahima_Chaudhari 1 +Fatma_Kusibeh 1 Lee_Baca 1 +Federico_Trillo 1 Jonathan_Woodgate 1 +Fernando_Alonso 1 Sam_Brownback 1 +Fernando_Sanz 1 Miranda_Otto 1 +Fernando_Sanz 1 Roy_Moore 1 +Flor_Montulo 2 Juan_Pablo_Montoya 1 +Francisco_Garcia 1 Marsha_Sharp 1 +Francois_Ozon 1 Makiya_Ali_Hassan 1 +Frank_Coraci 1 Tomoko_Hagiwara 2 +Frank_Van_Ecke 1 Tsutomu_Takebe 1 +Fred_Thompson 2 Helen_Alvare 1 +Fred_Thompson 2 Sterling_Hitchcock 1 +Fred_Thompson 3 Magda_Kertasz 1 +Garry_Trudeau 1 Pat_Riley 1 +Garry_Witherall 1 Howard_Stern 1 +Garry_Witherall 1 Ingrid_Betancourt 1 +Garry_Witherall 1 Martin_Keown 1 +Gary_Gero 1 Kim_Hong-gul 1 +Gary_Gero 1 Phil_Gramm 1 +Gavin_Degraw 1 Jeffrey_Archer 1 +Gene_Robinson 3 Martha_Bowen 2 +Georgia_Giddings 1 Mahima_Chaudhari 1 +Geovani_Lapentti 1 Rodney_Rempt 1 +Geovani_Lapentti 1 Sam_Brownback 1 +Gerard_de_Cortanze 1 Mark_Wahlberg 1 +Gian_Marco 2 Kevin_Stallings 2 +Giancarlo_Fisichella 1 Maria_Callas 1 +Gideon_Yago 1 Natalie_Williams 1 +Gideon_Yago 1 Paul_William_Hurley 1 +Glenn_Plummer 1 Maria_Garcia 1 +Grant_Hackett 1 Todd_Robbins 1 +Grant_Hackett 3 Milo_Djukanovic 3 +Gray_Davis 26 Karen_Lynn_Gorney 1 +Gregg_Popovich 3 Vernon_Forrest 1 +Gregor_Gysi 1 Tomoko_Hagiwara 1 +Guy_Ritchie 2 Herb_Dhaliwal 1 +Guy_Ritchie 2 William_Macy 1 +Hamid_Efendi 1 Jimmy_Carter 8 +Hamzah_Haz 2 Hilary_McKay 1 +Harald_Ringstorff 1 Pat_Riley 1 +Harald_Ringstorff 1 Romano_Prodi 6 +Heather_Chinnock 1 Jean-Francois_Lemounier 1 +Helen_Alvare 1 Milo_Djukanovic 1 +Henry_Castellanos 1 Pamela_Anderson 4 +Henry_Castellanos 1 Tommy_Shane_Steiner 1 +Herb_Dhaliwal 1 Hung_Wan-ting 1 +Hilary_McKay 1 Kevin_Millwood 1 +Howard_Stern 1 Maria_Callas 1 +Hugo_Chavez 33 Karen_Lynn_Gorney 1 +Hugo_Chavez 60 Steve_Shiver 1 +Imam_Samudra 1 Ivana_Trump 1 +Imelda_Marcos 1 Patty_Schnyder 4 +Jack_Smith 1 Mary_Jo_Myers 1 +James_Caan 2 Paul_Sarbanes 2 +James_Comey 1 Juan_Carlos_Morales 1 +James_Comey 1 Paul_William_Hurley 1 +Jamling_Norgay 1 Zico 2 +Jan_Pronk 1 Kim_Dong-hwa 1 +Janez_Drnovsek 1 Sterling_Hitchcock 1 +Janica_Kostelic 2 Yasushi_Akashi 1 +Janice_Abreu 1 Kevin_Sorbo 1 +Jeffrey_Ashby 1 Michael_Dell 1 +Jennifer_Garner 6 Mike_Duke 1 +Jennifer_Renee_Short 1 Taylor_Twellman 1 +Jerry_Seinfeld 1 Tim_Blake_Nelson 1 +Jerry_Tarkanian 1 Thomas_Rupprath 1 +Jessica_Lange 2 Sedigh_Barmak 1 +Jim_Freudenberg 1 Nigel_Redden 1 +Jim_Freudenberg 1 Tina_Pisnik 1 +Jim_Haslett 1 Tsutomu_Takebe 1 +Jim_Otto 1 Rafiq_Hariri 1 +Jimmy_Gurule 1 Terry_McAuliffe 1 +Jodie_Foster 3 Joe_Pantoliano 1 +John_Herrington 1 Luis_Ernesto_Derbez_Bautista 2 +John_Richardson 1 Yasushi_Akashi 1 +John_Snow 17 Se_Hyuk_Joo 1 +Jonathan_Arden 1 Joseph_Ralston 1 +Jorge_Castaneda 1 Robert_Fico 1 +Jose_Rosado 1 Micky_Arison 1 +Joseph_Blatter 1 Ronald_Kadish 1 +Joseph_Ralston 2 Juan_Pablo_Montoya 5 +Joseph_Ralston 2 Yoshiyuki_Kamei 1 +Juliette_Binoche 1 Matthew_Broderick 3 +Julio_De_Brun 1 Patty_Schnyder 1 +Julio_De_Brun 1 Vernon_Forrest 1 +Justin_Guarini 2 Prince_Edward 1 +Kate_Winslet 2 Mike_Duke 1 +Katie_Wagner 1 Stan_Kroenke 1 +Keith_Lowen 1 Robert_Evans 2 +Keith_Lowen 1 Silvia_Farina_Elia 2 +Ken_Loach 1 Taku_Yamasaki 1 +Kevin_Crane 1 Mike_Krzyzewski 2 +Kevin_Millwood 1 Mitchell_Crooks 1 +Kim_Clijsters 4 Martin_Short 1 +Kim_Hong-gul 1 Milo_Djukanovic 3 +Kim_Jong-Il 3 Rick_Reed 1 +Lance_Armstrong 9 Maria_Garcia 1 +Laurent_Jalabert 2 Vincent_Gallo 1 +Leon_LaPorte 1 Ted_Williams 1 +Leon_LaPorte 1 Tommy_Thompson 5 +Leonardo_Fernandez 1 Romano_Prodi 7 +Leticia_Dolera 1 Tom_Glavine 1 +Lorraine_Bracco 1 Momcilo_Perisic 1 +Luis_Ernesto_Derbez_Bautista 1 Tyra_Banks 1 +Maria_Burks 1 Todd_Parrott 1 +Mario_Lemieux 1 Stan_Kroenke 1 +Mark_Everson 1 Martin_Sheen 2 +Marquier_Montano_Contreras 1 SJ_Twu 1 +Marsha_Sharp 1 Steve_Shiver 1 +Martin_Cauchon 2 Vitali_Klitschko 3 +Martin_Hoellwarth 2 Mary_Katherine_Smart 1 +Martina_Hingis 1 Terry_McAuliffe 2 +Melina_Kanakaredes 1 Ornella_Muti 1 +Michael_Dell 1 Mike_Duke 1 +Michael_Dell 1 Nigel_Redden 1 +Michael_Richards 1 Silvia_Farina_Elia 3 +Milan_Kucan 1 Salman_Khan 1 +Nancy_Kerrigan 1 Sam_Brownback 1 +Naomi_Campbell 1 Tom_Ridge 16 +Nina_Jacobson 1 Portia_de_Rossi 1 +Noah_Wyle 3 Robbie_Coltrane 1 +Nora_Bendijo 1 William_Martin 2 +Nursultan_Nazarbayev 1 Robert_Bonner 1 +Pascal_Affi_Nguessan 1 Tom_Moss 1 +Pat_Summitt 1 Paul_Celluci 1 +Patty_Schnyder 3 Pernilla_Bjorn 1 +Patty_Schnyder 3 Prince_Philippe 1 +Patty_Schnyder 4 Ricardo_Lagos 25 +Pervez_Musharraf 3 Richard_Rodriguez 1 +Phil_Gramm 2 Stefan_Tafrov 1 +Rachel_Kempson 1 Zorica_Radovic 1 +Rachel_Roy 1 Steve_Shiver 1 +Richard_Fine 1 Richard_Rodriguez 1 +Rick_Reed 1 Ruth_Bader_Ginsburg 1 +Robbie_Naish 1 Zhong_Nanshan 1 +Robert_Bonner 2 Vincent_Brooks 2 +Robert_Downey_Jr 1 Tommy_Shane_Steiner 1 +Robert_Evans 1 Todd_Robbins 1 +Romeo_Gigli 1 Tom_Harkin 4 +Saeb_Erekat 1 Tom_Coverdale 2 +Se_Hyuk_Joo 1 Tom_Rouen 1 +Sergio_Garcia 2 Thomas_Watjen 1 +Simon_Yam 1 Terry_McAuliffe 3 +Simon_Yam 1 Tommy_Haas 5 +Stan_Kroenke 1 William_Hyde 1 +Steve_Ballmer 1 Tina_Pisnik 1 +Steve_Ballmer 2 Vincent_Gallo 3 +Steve_Shiver 1 Thomas_Rupprath 3 +Tina_Fey 1 Todd_Parrott 1 +Abdullah_Gul 1 6 +Abdullah_Gul 1 8 +Abdullah_Gul 7 14 +Abdullah_Gul 9 12 +Abdullah_Gul 9 15 +Adolfo_Rodriguez_Saa 1 2 +Adrien_Brody 2 3 +Adrien_Brody 2 12 +Adrien_Brody 5 10 +Adrien_Brody 7 8 +Al_Sharpton 1 4 +Al_Sharpton 2 4 +Al_Sharpton 2 7 +Al_Sharpton 3 4 +Alexandra_Stevenson 1 2 +Alexandra_Stevenson 1 3 +Alexandra_Vodjanikova 1 2 +Alicia_Silverstone 1 2 +Ana_Palacio 3 7 +Ana_Palacio 6 8 +Andre_Agassi 1 5 +Andre_Agassi 2 16 +Andre_Agassi 4 33 +Andre_Agassi 9 25 +Andre_Agassi 17 25 +Anna_Kournikova 2 3 +Anna_Kournikova 2 12 +Anna_Kournikova 3 8 +Anna_Kournikova 7 8 +Anna_Kournikova 7 11 +Annette_Lu 1 2 +Arnold_Palmer 1 3 +Arnold_Palmer 2 3 +Aron_Ralston 1 2 +Arturo_Gatti 2 3 +Bashar_Assad 1 3 +Bashar_Assad 2 4 +Bernardo_Segura 1 2 +Bill_Gates 3 16 +Bill_Gates 7 15 +Bill_Gates 11 13 +Bill_Gates 13 17 +Bo_Pelini 1 2 +Bob_Stoops 2 4 +Bob_Stoops 2 5 +Bobby_Robson 1 2 +Bode_Miller 1 2 +Caroline_Kennedy 1 3 +Caroline_Kennedy 2 3 +Catherine_Zeta-Jones 1 11 +Catherine_Zeta-Jones 4 9 +Celso_Amorim 1 2 +Chan_Gailey 1 2 +Chanda_Rubin 2 5 +Chanda_Rubin 4 5 +Charles_Bronson 1 2 +Charles_Kartman 1 2 +Charles_Schumer 1 2 +Chris_Rock 1 2 +Christine_Baumgartner 1 2 +Christine_Baumgartner 2 4 +Colin_Farrell 1 7 +Colin_Farrell 2 7 +Colin_Farrell 3 6 +Colin_Farrell 4 8 +Dalai_Lama 1 2 +Daniel_Radcliffe 1 4 +Daryl_Hannah 1 2 +David_Anderson 1 2 +David_Anderson 1 3 +David_Beckham 1 15 +David_Beckham 9 16 +David_Beckham 11 29 +David_Beckham 13 29 +David_Beckham 15 24 +David_Beckham 20 27 +David_Beckham 27 31 +Denzel_Washington 2 4 +Denzel_Washington 3 5 +Dianne_Feinstein 1 2 +Dianne_Feinstein 1 3 +Dianne_Feinstein 2 3 +Dick_Clark 1 2 +Dick_Clark 1 3 +Donald_Fehr 1 3 +Donald_Fehr 1 4 +Donald_Fehr 2 3 +Donald_Fehr 3 4 +Dwayne_Johnson 1 2 +Ed_Rosenthal 1 2 +Erik_Morales 1 2 +Erik_Morales 2 3 +Evan_Rachel_Wood 2 3 +Evander_Holyfield 1 2 +Eve_Pelletier 1 2 +Farouk_al-Sharaa 1 2 +Farouk_al-Sharaa 1 3 +Farouk_al-Sharaa 2 3 +Frank_Cassell 1 2 +Frank_Cassell 1 3 +Fujio_Cho 1 4 +Fujio_Cho 2 5 +Fujio_Cho 3 5 +Fujio_Cho 4 6 +Fujio_Cho 5 6 +Gao_Qiang 1 2 +Geoff_Hoon 1 2 +Geoff_Hoon 3 4 +Geoff_Hoon 4 5 +George_Brumley 1 2 +George_Papandreou 1 2 +George_Papandreou 1 3 +George_Papandreou 1 4 +George_Papandreou 3 4 +Gloria_Allred 1 2 +Greg_Ostertag 1 2 +Greg_Owen 1 2 +Hanan_Ashrawi 1 2 +Harry_Kalas 1 2 +Hayley_Tullett 1 2 +Howard_Dean 1 3 +Howard_Dean 1 7 +Howard_Dean 2 8 +Howard_Dean 4 5 +Howard_Dean 5 12 +Hu_Jintao 4 14 +Hu_Jintao 5 12 +Hu_Jintao 5 15 +Hu_Jintao 11 13 +Ian_McKellen 1 3 +Ibrahim_Jaafari 1 2 +Isabella_Rossellini 1 3 +Ishaq_Shahryar 1 2 +Ismail_Merchant 1 2 +Jacques_Chirac 2 43 +Jacques_Chirac 19 28 +James_Cameron 1 3 +James_McGreevey 1 3 +James_McGreevey 1 4 +Jason_Alexander 1 2 +Jason_Kidd 2 8 +Jason_Kidd 3 8 +Jason_Kidd 6 8 +Jelena_Dokic 1 2 +Jelena_Dokic 2 4 +Jelena_Dokic 2 8 +Jelena_Dokic 3 4 +Jelena_Dokic 4 5 +Jelena_Dokic 4 7 +Jelena_Dokic 4 8 +Jelena_Dokic 5 6 +Jelena_Dokic 7 8 +Jennifer_Connelly 1 3 +Jennifer_Connelly 1 4 +Jennifer_Connelly 2 4 +Jennifer_Connelly 3 4 +Jennifer_Keller 1 4 +Jennifer_Keller 3 4 +Jennifer_Reilly 1 2 +Jennifer_Thompson 1 2 +Jim_Hahn 1 3 +Jim_Hahn 1 4 +Jim_Hahn 3 4 +Johnny_Carson 1 2 +Jorge_Batlle 1 2 +Jorge_Batlle 1 3 +Jorge_Rodolfo_Canicoba_Corral 1 2 +Juan_Ignacio_Chela 1 2 +Juan_Ignacio_Chela 1 3 +Juan_Ignacio_Chela 2 3 +Justin_Gatlin 1 2 +Karen_Mok 1 2 +Katie_Harman 1 2 +Katie_Harman 1 3 +Katie_Harman 2 3 +Kenneth_Branagh 1 2 +Kevin_Costner 1 3 +Kevin_Costner 2 6 +Kevin_Costner 2 8 +Kevin_Costner 6 8 +Larry_Lindsey 1 2 +Leonardo_DiCaprio 3 7 +Leonardo_DiCaprio 5 6 +Leonardo_DiCaprio 6 8 +Lindsay_Davenport 2 10 +Lindsay_Davenport 5 9 +Lisa_Ling 1 2 +Lucy_Liu 2 4 +Lucy_Liu 2 5 +Lucy_Liu 3 5 +Manfred_Stolpe 1 2 +Maria_Luisa_Mendonca 1 2 +Mario_Kreutzberger 1 2 +Melissa_Etheridge 1 2 +Michael_Jordan 1 2 +Michael_Jordan 1 3 +Michael_Jordan 2 4 +Mikhail_Youzhny 1 2 +Mitchell_Daniels 1 2 +Mitchell_Daniels 1 4 +Mitchell_Daniels 3 4 +Mohammad_Khatami 1 5 +Mohammad_Khatami 3 9 +Mohammad_Khatami 8 10 +Monique_Garbrecht-Enfeldt 1 3 +Natalie_Maines 2 5 +Natalie_Maines 4 5 +Nelson_Mandela 1 3 +Norodom_Sihanouk 1 3 +Norodom_Sihanouk 2 3 +Osama_bin_Laden 2 4 +Osama_bin_Laden 3 4 +Oxana_Fedorova 1 2 +Peter_Bacanovic 1 2 +Philippe_Noiret 1 2 +Placido_Domingo 1 2 +Placido_Domingo 2 3 +Prince_William 1 2 +Princess_Aiko 1 2 +Priscilla_Presley 1 2 +Richard_Haass 1 2 +Richard_Shelby 1 2 +Rick_Barnes 1 2 +Rick_Barnes 1 3 +Rick_Dinse 1 3 +Rick_Dinse 2 3 +Rio_Ferdinand 1 2 +Rita_Grande 1 2 +Rita_Grande 2 3 +Robin_McLaurin_Williams 1 2 +Saddam_Hussein 2 4 +Saddam_Hussein 2 10 +Saddam_Hussein 3 11 +Saddam_Hussein 6 9 +Sally_Field 1 2 +Salma_Hayek 1 10 +Salma_Hayek 1 11 +Sandra_Bullock 1 3 +Sandra_Bullock 1 4 +Sandra_Bullock 2 4 +Sandra_Bullock 3 4 +Sarah_Hughes 1 6 +Sarah_Hughes 2 5 +Sarah_Hughes 3 5 +Sarah_Hughes 4 6 +Sharon_Frey 1 2 +Silvio_Berlusconi 6 30 +Silvio_Berlusconi 8 23 +Silvio_Berlusconi 13 21 +Silvio_Berlusconi 17 29 +Silvio_Berlusconi 18 25 +Silvio_Berlusconi 23 28 +Silvio_Berlusconi 24 30 +Silvio_Berlusconi 26 32 +Spencer_Abraham 3 17 +Steffi_Graf 2 5 +Steffi_Graf 3 4 +Steven_Spielberg 1 5 +Steven_Spielberg 1 6 +Steven_Spielberg 3 6 +Steven_Spielberg 4 6 +Steven_Spielberg 4 7 +Steven_Spielberg 6 7 +Theresa_May 1 2 +Theresa_May 2 3 +Tippi_Hedren 1 2 +Todd_Haynes 1 2 +Todd_Haynes 1 3 +Todd_Haynes 1 4 +Todd_Haynes 2 4 +Tony_Blair 18 86 +Tony_Blair 32 48 +Tony_Blair 32 110 +Tony_Blair 53 102 +Tony_Blair 91 134 +Tony_Blair 92 133 +Tony_Blair 105 121 +Vicente_Fox 4 30 +Vicente_Fox 9 16 +Vicente_Fox 11 17 +Vicente_Fox 15 30 +Vojislav_Kostunica 1 3 +Vojislav_Kostunica 2 4 +Vojislav_Kostunica 4 7 +Vojislav_Kostunica 5 7 +Walter_Mondale 1 6 +Walter_Mondale 1 7 +Walter_Mondale 5 7 +Walter_Mondale 6 8 +Walter_Mondale 7 9 +Wang_Yingfan 2 3 +William_Bratton 1 3 +William_Donaldson 1 5 +William_Donaldson 3 5 +William_Donaldson 4 6 +Xavier_Malisse 1 3 +Xavier_Malisse 3 4 +Xavier_Malisse 3 5 +Yevgeny_Kafelnikov 1 4 +Zarai_Toledo 1 2 +Adolfo_Rodriguez_Saa 1 Dave_Odom 1 +Adolfo_Rodriguez_Saa 1 Nancy_Powell 1 +Adrien_Brody 3 Damon_Dash 1 +Ahmed_Ahmed 1 Mike_Smith 1 +Al_Sharpton 5 Cole_Chapman 1 +Alberto_Sordi 1 James_Cameron 2 +Alejandro_Lerner 1 Jesper_Parnevik 1 +Alejandro_Lerner 1 TA_McLendon 1 +Alex_Penelas 1 Robin_Johansen 1 +Alex_Penelas 2 Gretchen_Mol 1 +Alexandra_Stevenson 3 Ronde_Barber 1 +Ali_Adbul_Karim_Madani 1 Carey_Lowell 1 +Ali_Mohammed_Maher 1 Isabela_Moraes 1 +Ali_Mohammed_Maher 1 Isidro_Pastor 1 +Alicia_Silverstone 2 Jennifer_Thompson 1 +Alicia_Silverstone 2 Jonathan_Tiomkin 1 +Alicia_Silverstone 2 Serge_Tchuruk 1 +Andre_Agassi 31 Boris_Jordan 1 +Andre_Techine 1 Lloyd_Richards 1 +Andy_Benes 1 Doug_Moe 1 +Andy_Benes 1 Mitar_Rasevic 1 +Andy_Bryant 1 Mike_Slive 1 +Anna_Kournikova 8 Colin_Farrell 1 +Anne_ONeil 1 Sophie 1 +Anthony_Lee_Johnson 1 Dave_Williams 1 +Arnold_Palmer 1 Stephanie_Zimbalist 1 +Arnold_Palmer 3 Charles_Kartman 1 +Aron_Ralston 1 Bill_Fennelly 1 +Bart_Hendricks 1 Philippe_Noiret 1 +Ben_Kingsley 1 Daryl_Hannah 1 +Ben_Kingsley 1 Jean_Nagel 1 +Ben_Kingsley 1 Perry_Farrell 1 +Benjamin_Neulander 1 Robin_Tunney 1 +Bernardo_Segura 1 Debra_Shank 1 +Bernardo_Segura 2 Douglas_Paal 1 +Bernardo_Segura 2 Jeanette_Stauffer 1 +Bill_Pryor 1 Ronde_Barber 1 +Bob_Melvin 1 Richard_Hamilton 1 +Bob_Stoops 3 Wayne_Allard 1 +Bobby_Jackson 1 Bruce_Gebhardt 1 +Bobby_Jackson 1 JP_Suarez 1 +Bobby_Jackson 1 Madge_Overhouse 1 +Bobby_Jackson 1 Nicola_Wells 1 +Bobby_Robson 1 Rollie_Massimino 1 +Bode_Miller 1 David_Howard 1 +Bode_Miller 1 Steffi_Graf 1 +Boris_Jordan 1 Kim_Weeks 1 +Boris_Jordan 1 Vincent_Sombrotto 1 +Brad_Alexander_Smith 1 Jason_Alexander 2 +Brad_Alexander_Smith 1 Kate_Lee 1 +Brad_Alexander_Smith 1 Tatiana_Gratcheva 1 +Brandon_Webb 1 Helmut_Panke 1 +Brandon_Webb 1 Larry_Hahn 1 +Brandon_Webb 1 Ryan_Leaf 1 +Brennon_Leighton 1 Laurel_Clark 1 +Brett_Hawke 1 Teri_Files 1 +Bruce_Gebhardt 1 Jean-Luc_Bideau 1 +Bruce_Gebhardt 1 Masao_Azuma 1 +Bryan_Cooley 1 Katie_Harman 3 +Bryan_Murray 1 Eric_Schacht 1 +Bryan_Murray 1 Mikhail_Youzhny 1 +Bryan_Murray 1 Mitchell_Daniels 2 +Camille_Colvin 1 Irina_Yatchenko 1 +Carey_Lowell 1 Charlie_Coles 1 +Carl_Pope 1 Larry_Hahn 1 +Carolina_Barco 1 Daryl_Hannah 1 +Carolina_Barco 1 Lindsay_Davenport 22 +Carolina_Barco 1 Norodom_Sihanouk 1 +Caroline_Kennedy 3 Henry_Suazo 1 +Catherine_Bell 1 Guillaume_Cannet 1 +Catherine_Zeta-Jones 10 Pier_Ferdinando_Casini 1 +Catherine_Zeta-Jones 11 Hayley_Tullett 2 +Celso_Amorim 2 Juljia_Vysotskij 1 +Celso_Amorim 2 Kevin_Costner 1 +Celso_Amorim 2 Thomas_Stewart 1 +Chanda_Rubin 3 Richard_Tubb 1 +Chante_Jawan_Mallard 1 Stephen_Glassroth 1 +Chante_Jawan_Mallard 1 Vicente_Fox 23 +Charles_Bronson 2 Hu_Jintao 4 +Charles_Kartman 1 Debra_Shank 1 +Charles_Schumer 2 James_Watt 1 +Charles_Schumer 2 Justin_Gatlin 2 +Charles_Schumer 2 Lionel_Chalmers 1 +Chris_Rock 2 Dalai_Lama 1 +Christine_Baumgartner 2 Vincent_Sombrotto 1 +Christine_Baumgartner 3 Lars_Burgsmuller 1 +Chyung_Dai-chul 1 Dan_Reeves 1 +Cindy_Moll 1 Daniel_Radcliffe 3 +Cindy_Moll 1 James_McPherson 1 +Cindy_Moll 1 Robin_Johansen 1 +Claire_De_Gryse 1 Deb_Santos 1 +Claire_Tomalin 1 Steve_Case 1 +Clay_Campbell 1 Leonardo_DiCaprio 2 +Cole_Chapman 1 Marisol_Martinez_Sambran 1 +Colin_Phillips 1 Ian_McKellen 3 +Colin_Phillips 1 Lloyd_Richards 1 +Colin_Prescot 1 Laurie_Hobbs 1 +Connie_Freydell 1 Tommy_Maddox 1 +Craig_OClair 1 Steve_Avery 1 +Damarius_Bilbo 1 Juljia_Vysotskij 1 +Damon_Dash 1 Henry_Suazo 1 +Daniel_Comisso_Urdaneta 1 Larry_Ralston 1 +Daniel_Radcliffe 3 Diane_Green 2 +Daniel_Radcliffe 3 Rick_Barnes 2 +Daniel_Radcliffe 3 Vicente_Fox 32 +Dany_Heatley 1 Richard_Parsons 1 +Dany_Heatley 1 Terry_Hoeppner 1 +Dave_Odom 1 Maritza_Macias_Furano 1 +David_Beckham 22 Laura_Morante 1 +David_Duval 1 Philippe_Noiret 2 +David_Ho 1 Doug_Moe 1 +David_Ho 1 Erskine_Bowles 1 +David_Ho 1 Evander_Holyfield 1 +David_Howard 1 Hugo_Conte 1 +Denzel_Washington 4 Henry_Suazo 1 +Des_Brown 1 Eliott_Spitzer 1 +Des_Brown 1 Svetislav_Pesic 1 +Diana_Ross 1 Rick_Rickert 1 +Dianne_Feinstein 3 Dwayne_Johnson 2 +Dick_Clark 3 Manfred_Stolpe 1 +Dino_Risi 1 Gabriel_Farhi 1 +Doc_Rivers 1 Melissa_Joan_Hart 1 +Dominick_Dunne 1 Nicoletta_Braschi 1 +Don_Flanagan 1 Ed_Sullivan 1 +Don_Flanagan 1 Ratna_Sari_Dewi_Sukarno 1 +Donald_Keyser 1 Salma_Hayek 2 +Douglas_Paal 1 Ed_Sullivan 1 +Douglas_Paal 1 Richard_Hamilton 1 +Dusty_Baker 1 Saddam_Hussein 2 +Ed_Rendell 1 Jamie_Dimon 1 +Ed_Rendell 1 Richard_Cohen 1 +Eddie_Murray 1 Kevin_Costner 7 +Elaine_Stritch 1 Richard_Parsons 1 +Elena_Tihomirova 1 Mike_Slive 1 +Elena_Tihomirova 1 Mohammed_Abu_Sharia 1 +Eric_Schacht 1 Jennifer_Keller 2 +Erik_Morales 3 Werner_Schlager 1 +Erin_Brockovich 1 Henry_Suazo 1 +Erin_Brockovich 1 Mike_Carona 1 +Evan_Rachel_Wood 1 Sharon_Frey 1 +Evander_Holyfield 1 Julio_Rossi 1 +Evander_Holyfield 1 Michael_Jordan 3 +Evander_Holyfield 1 Mike_Carona 1 +Fatmir_Limaj 1 Todd_Haynes 4 +Felicity_Huffman 1 Nelson_Acosta 1 +Felipe_De_Borbon 1 Jose_Bove 1 +Frank_Cassell 2 Jesper_Parnevik 1 +Frank_Cassell 3 Ronde_Barber 1 +Fred_Durst 1 Marisol_Martinez_Sambran 1 +Fred_Durst 1 Shavon_Earp 1 +Fujio_Cho 1 Javier_Vargas 1 +Fujio_Cho 2 Tatsuya_Fuji 1 +Fujio_Cho 3 Michael_Brandon 1 +Fujio_Cho 6 Olene_Walker 1 +Gabriel_Farhi 1 Lars_Burgsmuller 1 +Gao_Qiang 2 Kenneth_Brill 1 +Gary_Stevens 1 Wanda_Ilene_Barzee 1 +Geoff_Hoon 1 Osama_bin_Laden 3 +Geoff_Hoon 3 Richard_Tubb 1 +Geoff_Hoon 5 Thomas_Stewart 1 +George_Brumley 1 Manfred_Stolpe 1 +George_Brumley 1 Ryan_Leaf 1 +George_Brumley 2 Sergei_Yushenkov 1 +George_Papandreou 2 Tatiana_Panova 1 +George_Papandreou 4 Larry_Lindsey 2 +George_Papandreou 4 Paul_LeClerc 1 +Ghassan_Elashi 1 Rick_Rickert 1 +Gloria_Allred 2 Roman_Tam 1 +Greg_Kinsey 1 Priscilla_Presley 1 +Greg_Ostertag 2 Kevin_Costner 6 +Greg_Owen 1 Mark_Sisk 1 +Greg_Owen 2 Mike_Smith 1 +Gretchen_Mol 1 Isabella_Rossellini 2 +Gunilla_Backman 1 Kathleen_Abernathy 1 +Hanan_Ashrawi 1 Mario_Kreutzberger 2 +Hans_Leistritz 1 Saddam_Hussein 18 +Hans_Leistritz 1 Sue_Grafton 1 +Hans_Peter_Briegel 1 Theresa_May 2 +Harriet_Lessy 1 Mike_Carona 1 +Harry_Kalas 2 Joe_Strummer 1 +Harry_Kalas 2 Lewis_Booth 1 +Hayley_Tullett 1 Ian_McKellen 1 +Henry_Suazo 1 Ratna_Sari_Dewi_Sukarno 1 +Hernan_Crespo 1 Ronde_Barber 1 +Hiroyuki_Yoshino 1 Rollie_Massimino 1 +Hu_Jintao 6 Jerome_Jenkins 1 +Hugh_Jessiman 1 Solomon_Passy 1 +Hugh_Jessiman 1 TA_McLendon 1 +Hugo_Conte 1 Laurent_Woulzy 1 +Irina_Yatchenko 1 Katie_Harman 2 +Isidro_Pastor 1 Mark_Butcher 1 +Ismail_Merchant 1 Mack_Brown 2 +Ismail_Merchant 2 Kathleen_Abernathy 1 +Ivan_Helguera 1 William_Donaldson 4 +Iveta_Benesova 1 Jacques_Chirac 50 +Jacques_Chirac 20 Tzipora_Obziler 1 +Jacques_Chirac 35 Richard_Shelby 2 +James_Cameron 1 Richard_Shelby 2 +James_McGreevey 1 Katie_Harman 2 +James_McGreevey 1 Tzipora_Obziler 1 +James_Sensenbrenner 1 Jimmy_Smits 1 +James_Sensenbrenner 1 Maria_Shkolnikova 1 +James_Watt 1 Priscilla_Presley 2 +Jamie_Dimon 1 Robert_Korzeniowski 1 +Jason_Alexander 1 Mike_Maroth 1 +Jason_Alexander 1 Wayne_Allard 1 +Jason_Kidd 7 Louisa_Baileche 1 +Javier_Vargas 1 Jesse_Helms 1 +Jeanette_Gray 1 Keith_Fotta 1 +Jeanette_Stauffer 1 Marie-Josee_Croze 1 +Jeffrey_Donaldson 1 Leonardo_DiCaprio 8 +Jeffrey_Donaldson 1 Peter_Bacanovic 1 +Jennifer_Connelly 4 Martin_Bandier 1 +Jesper_Parnevik 1 Justin_Gatlin 2 +Jimmy_Smits 1 Robin_Johansen 1 +Jimmy_Smits 1 Shane_Hmiel 1 +Joe_Crede 1 Tom_DeLay 1 +John_Burkett 1 Leonardo_DiCaprio 2 +John_McKay 1 Mary_Sue_Coleman 1 +Johnny_Hallyday 1 Olene_Walker 1 +Jonathan_Tiomkin 1 Rita_Grande 2 +Jorge_Rodolfo_Canicoba_Corral 1 Wayne_Allard 1 +Jose_Bove 1 Juergen_Trittin 1 +Juljia_Vysotskij 1 Patrick_Coleman 1 +Kate_Lee 1 Toshimitsu_Motegi 1 +Kathie_Louise_Saunders 1 Roberto_Lavagna 1 +Kathie_Louise_Saunders 1 Rosalyn_Carter 1 +Kenneth_Branagh 1 Pinar_del_Rio 1 +Kevin_Costner 5 Stephen_Glassroth 1 +Khader_Rashid_Rahim 1 Richard_Cohen 1 +Khalid_Khannouchi 1 Werner_Schlager 1 +Kim_Su_Nam 1 Todd_Haynes 4 +Kim_Weeks 1 Marie-Josee_Croze 1 +Kimora_Lee 1 Robin_Johansen 1 +Kurt_Suzuki 1 Scott_Hoch 1 +Larry_Lindsey 1 Wanda_Ilene_Barzee 1 +Larry_Ralston 1 Richard_Carl 1 +Larry_Ralston 1 William_Donaldson 6 +Larry_Ralston 1 Willie_Wilson 1 +Laurel_Clark 1 Laurent_Woulzy 1 +Laurel_Clark 1 Priscilla_Presley 1 +Leandrinho_Barbosa 1 Melissa_Joan_Hart 1 +Leandrinho_Barbosa 1 Olene_Walker 1 +Leon_Barmore 1 Sachin_Tendulkar 1 +Leon_Barmore 1 Yusaku_Miyazato 1 +Leonard_Schrank 1 Marie_Haghal 1 +Leonard_Schrank 1 Rick_Rickert 1 +Linda_Franklin 1 Melissa_Joan_Hart 1 +Lindsay_Davenport 1 Martin_Burnham 1 +Lindsay_Davenport 17 Oxana_Fedorova 3 +Lionel_Chalmers 1 Mohammad_Fares 1 +Lisa_Ling 2 Mike_Maroth 1 +Lucy_Liu 3 Prince_William 1 +Maria_Luisa_Mendonca 2 Stephane_Delajoux 1 +Maria_Luisa_Mendonca 2 Terunobu_Maeda 1 +Maria_Shkolnikova 1 Martin_Rodriguez 1 +Mariana_Ohata 1 Xavier_Malisse 3 +Mario_Kreutzberger 1 Shavon_Earp 1 +Mario_Kreutzberger 2 Raul_Cubas 1 +Maritza_Macias_Furano 1 Qusai_Hussein 1 +Mark_Butcher 1 Scott_Hoch 1 +Mark_Sisk 1 Mehdi_Baala 1 +Mark_Sisk 1 Stephen_Swindal 1 +Masao_Azuma 1 Mikhail_Kalashnikov 1 +McGuire_Gibson 1 Richard_Haass 1 +Mehdi_Baala 1 Steve_Avery 1 +Melissa_Joan_Hart 1 Mikhail_Kalashnikov 1 +Melissa_Joan_Hart 1 Sue_Grafton 1 +Michael_Brandon 1 Toshimitsu_Motegi 1 +Michael_Shelby 1 Olene_Walker 1 +Michel_Kratochvil 1 Sheldon_Silver 1 +Mike_Eskew 1 Zarai_Toledo 1 +Mike_OConnell 1 Roberto_Lavagna 1 +Mike_Sherman 1 Natalie_Maines 3 +Mike_Sherman 1 Paige_Fitzgerald 1 +Milt_Heflin 1 Pier_Ferdinando_Casini 1 +Nicholoas_DiMarzio 1 Richard_Parsons 1 +Nikki_McKibbin 1 Stephen_Swindal 1 +Nikki_McKibbin 1 Steven_Tyler 1 +Osama_Al_Baz 1 Patricia_Wartusch 1 +Osama_Al_Baz 1 Sandra_Day_OConner 1 +Patricia_Phillips 1 Thierry_Falise 2 +Paula_Locke 1 Teri_Files 1 +Peter_Bacanovic 2 Robert_Korzeniowski 1 +Pier_Ferdinando_Casini 1 Ratna_Sari_Dewi_Sukarno 1 +Priscilla_Presley 2 Raul_Rivero 1 +Raaf_Schefter 1 Rick_Dinse 2 +Richard_Sterner 1 Steven_Spielberg 2 +Rick_Dinse 2 Shavon_Earp 1 +Roberto_Lavagna 1 Sandra_Bullock 4 +Rod_Stewart 1 Steffi_Graf 2 +Rodolfo_Abalos 1 Thierry_Falise 3 +Roman_Tam 1 Zalmay_Khalilzad 1 +Scott_Hoch 1 Thomas_Stewart 1 +Seymour_Cassell 1 Todd_Haynes 4 +Shaun_Rusling 1 Vincent_Sombrotto 1 +Sheldon_Silver 1 Xavier_Malisse 3 +Stephane_Delajoux 1 Tristan_Gretzky 1 +Takahiro_Mori 1 Tony_Blair 64 +Thierry_Falise 2 Werner_Schlager 1 +Aaron_Sorkin 1 2 +Abdullah 1 3 +Abdullah 2 3 +Abdullah 2 4 +Abdullah 3 4 +Abid_Hamid_Mahmud_Al-Tikriti 1 2 +Abid_Hamid_Mahmud_Al-Tikriti 1 3 +Alan_Ball 1 2 +Albert_Costa 3 5 +Albert_Costa 5 6 +Ali_Khamenei 1 2 +Ali_Khamenei 1 3 +Amelia_Vega 1 7 +Amelia_Vega 2 7 +Amelia_Vega 3 5 +Antonio_Banderas 1 5 +Antonio_Banderas 2 5 +Antonio_Banderas 3 5 +Arye_Mekel 1 2 +Azra_Akin 1 3 +Azra_Akin 1 4 +Bernard_Landry 2 3 +Bernard_Landry 2 4 +Bernard_Landry 3 4 +Biljana_Plavsic 1 2 +Biljana_Plavsic 1 3 +Biljana_Plavsic 2 3 +Bob_Hope 1 5 +Bob_Hope 3 6 +Bob_Hope 4 5 +Bob_Hope 4 7 +Bridget_Fonda 1 3 +Bridget_Fonda 2 3 +Carlos_Ruiz 1 3 +Carlos_Ruiz 2 3 +Carly_Fiorina 1 2 +Carly_Fiorina 1 3 +Cherie_Blair 1 3 +Cherie_Blair 1 4 +Cherie_Blair 2 4 +Cherie_Blair 3 4 +Chuck_Yeager 1 2 +Chung_Mong-joon 1 2 +Claire_Hentzen 1 2 +Claire_Leger 1 2 +Darren_Clarke 1 2 +David_Caraway 1 2 +David_Leahy 1 2 +Dick_Cheney 2 14 +Dick_Cheney 3 11 +Dick_Cheney 3 12 +Dick_Cheney 8 10 +Dino_de_Laurentis 1 2 +Don_Nickles 1 2 +Doris_Schroeder 1 3 +Doris_Schroeder 2 4 +Eduardo_Duhalde 2 3 +Eduardo_Duhalde 2 12 +Eduardo_Duhalde 3 4 +Eduardo_Duhalde 6 13 +Eduardo_Duhalde 7 10 +Eduardo_Duhalde 10 12 +Edwin_Edwards 1 3 +Eric_Rosser 1 2 +Ernie_Eves 1 2 +Ernie_Fletcher 1 2 +Eva_Dimas 1 2 +Fernando_Vargas 1 2 +Fernando_Vargas 1 3 +Fernando_Vargas 1 4 +Fernando_Vargas 2 3 +Fernando_Vargas 3 4 +Frank_Lautenberg 1 2 +George_HW_Bush 1 3 +George_HW_Bush 2 13 +George_HW_Bush 4 6 +George_HW_Bush 5 8 +George_HW_Bush 9 10 +Gisele_Bundchen 1 2 +Glafcos_Clerides 1 2 +Glafcos_Clerides 1 3 +Glafcos_Clerides 2 3 +Glafcos_Clerides 2 4 +Greg_Gilbert 1 2 +Hashim_Thaci 1 2 +Hassan_Wirajuda 1 2 +Hector_Babenco 1 3 +Hector_Babenco 2 3 +Hideki_Matsui 1 2 +Hillary_Clinton 1 10 +Hillary_Clinton 1 14 +Hillary_Clinton 4 5 +Hillary_Clinton 7 10 +Hillary_Clinton 11 13 +Hisao_Oguchi 1 2 +Hitomi_Soga 1 4 +Hitomi_Soga 1 5 +Hitomi_Soga 2 3 +Hitomi_Soga 3 4 +JJ_Redick 1 2 +Jean_Brumley 1 2 +Jean_Carnahan 1 2 +Jeremy_Shockey 1 2 +Jerry_Regier 1 3 +Jerry_Regier 2 3 +Jerry_Springer 1 2 +Jerry_Springer 1 4 +Jessica_Lynch 1 2 +Jiang_Zemin 1 17 +Jiang_Zemin 2 3 +Jiang_Zemin 4 10 +Jiang_Zemin 7 9 +Jiang_Zemin 15 18 +Jim_Harrick 1 2 +Jiri_Novak 3 4 +Jiri_Novak 3 8 +Jiri_Novak 3 9 +Jiri_Novak 5 6 +Jiri_Novak 5 10 +Jiri_Novak 5 11 +Jiri_Novak 7 10 +Joe_Torre 1 3 +Joe_Torre 2 4 +John_Wolf 1 2 +Jonathan_Edwards 1 6 +Jonathan_Edwards 2 7 +Jonathan_Edwards 3 6 +Jonathan_Edwards 5 6 +Jonathan_Edwards 5 7 +Jonathan_Edwards 6 8 +Jose_Manuel_Durao_Barroso 1 2 +Jose_Manuel_Durao_Barroso 1 6 +Jose_Manuel_Durao_Barroso 2 5 +Jose_Manuel_Durao_Barroso 3 5 +Jose_Manuel_Durao_Barroso 4 5 +Joseph_Biden 1 4 +Joseph_Biden 2 3 +Joseph_Biden 3 4 +Judi_Dench 1 2 +Judy_Genshaft 1 2 +Keanu_Reeves 1 9 +Keanu_Reeves 4 7 +Keanu_Reeves 6 9 +Keanu_Reeves 6 10 +Keanu_Reeves 7 12 +Keira_Knightley 1 2 +Ken_Watanabe 1 2 +Kieran_Prendergast 1 2 +King_Abdullah_II 1 3 +King_Abdullah_II 1 4 +King_Abdullah_II 2 5 +King_Abdullah_II 3 4 +Kirk_Ferentz 1 2 +Kurt_Busch 1 2 +Larry_Bowa 1 2 +Larry_Thompson 1 3 +Larry_Thompson 1 4 +Larry_Thompson 2 4 +Laura_Bush 1 11 +Laura_Bush 21 24 +Laura_Bush 21 39 +Laura_Bush 26 29 +Lauren_Hutton 1 2 +Leslie_Ann_Woodward 1 2 +Leslie_Moonves 1 2 +Lyle_Vanclief 1 2 +Magui_Serna 1 2 +Makhdoom_Amin_Fahim 1 3 +Makhdoom_Amin_Fahim 2 3 +Marc-Andre_Fleury 1 2 +Marco_Antonio_Barrera 1 5 +Marco_Antonio_Barrera 5 6 +Marie-Reine_Le_Gougne 1 2 +Marieta_Chrousala 1 2 +Marieta_Chrousala 1 3 +Marieta_Chrousala 2 3 +Marina_Anissina 1 2 +Mark_Hamister 1 2 +Martha_Stewart 1 3 +Martha_Stewart 1 5 +Martha_Stewart 2 3 +Martha_Stewart 2 5 +Martha_Stewart 3 4 +Martin_Brodeur 1 2 +Martin_McCauley 1 2 +Martin_Verkerk 1 2 +Martin_Verkerk 1 3 +Martin_Verkerk 2 3 +Martina_McBride 1 3 +Martina_McBride 2 4 +Matt_Doherty 1 2 +Matt_Doherty 1 3 +Matt_Doherty 2 3 +Matthew_Perry 1 6 +Matthew_Perry 2 7 +Matthew_Perry 3 4 +Megawati_Sukarnoputri 6 23 +Megawati_Sukarnoputri 7 22 +Megawati_Sukarnoputri 10 15 +Megawati_Sukarnoputri 11 24 +Megawati_Sukarnoputri 20 24 +Megawati_Sukarnoputri 20 30 +Meghann_Shaughnessy 1 2 +Michael_Capellas 1 2 +Michael_Sullivan 1 2 +Mike_Brey 1 2 +Naji_Sabri 1 6 +Naji_Sabri 2 4 +Naji_Sabri 2 7 +Naji_Sabri 3 8 +Naji_Sabri 6 7 +Nanni_Moretti 1 2 +Nastassia_Kinski 1 2 +Natalie_Coughlin 2 3 +Natalie_Coughlin 4 6 +Natalie_Coughlin 5 6 +Norah_Jones 3 15 +Norah_Jones 4 12 +Norah_Jones 7 15 +Norah_Jones 9 15 +Norah_Jones 11 12 +Norm_Coleman 5 7 +Oscar_De_La_Hoya 1 3 +Oscar_De_La_Hoya 2 6 +Oscar_De_La_Hoya 2 7 +Pascal_Lamy 1 2 +Pat_Burns 1 2 +Paul_McCartney 3 4 +Paul_McCartney 3 5 +Paul_Wellstone 1 2 +Paul_Wellstone 1 3 +Paul_Wellstone 2 3 +Penelope_Cruz 1 2 +Penelope_Cruz 1 3 +Pete_Rose 1 2 +Prince_Harry 1 2 +Prince_Harry 2 3 +Princess_Caroline 1 2 +Princess_Caroline 2 5 +Raquel_Welch 1 2 +Reggie_Miller 1 2 +Renee_Zellweger 2 13 +Renee_Zellweger 2 16 +Renee_Zellweger 3 16 +Ricardo_Sanchez 1 6 +Ricardo_Sanchez 2 4 +Richard_Butler 1 2 +Rubens_Barrichello 2 3 +Rubens_Barrichello 4 5 +Rubens_Barrichello 4 8 +Rubens_Barrichello 6 11 +Rubens_Barrichello 9 12 +Samira_Makhmalbaf 1 2 +Samuel_Waksal 1 2 +Samuel_Waksal 1 3 +Samuel_Waksal 1 4 +Samuel_Waksal 2 4 +Scott_McClellan 1 3 +Scott_McClellan 2 4 +Scott_McClellan 2 5 +Scott_Rudin 1 2 +Shane_Warne 1 2 +Sheila_Wellstone 1 2 +Sheryl_Crow 1 3 +Silvan_Shalom 2 3 +Silvan_Shalom 2 4 +Silvan_Shalom 2 6 +Silvan_Shalom 3 4 +Silvan_Shalom 3 6 +Steve_Mariucci 1 3 +Steve_Mariucci 2 3 +Steven_Seagal 1 2 +Taha_Yassin_Ramadan 3 12 +Taha_Yassin_Ramadan 7 13 +Tammy_Lynn_Michaels 1 2 +Theodore_Tweed_Roosevelt 1 2 +Theodore_Tweed_Roosevelt 1 3 +Theodore_Tweed_Roosevelt 2 3 +Tom_Crean 1 2 +Tom_Crean 3 5 +Tom_Crean 4 5 +Tony_Bennett 1 3 +Torri_Edwards 1 2 +Tung_Chee-hwa 1 2 +Tung_Chee-hwa 1 8 +Tung_Chee-hwa 2 8 +Vidar_Helgesen 1 2 +Warren_Buffett 1 2 +Warren_Buffett 1 3 +Wen_Jiabao 1 12 +Wen_Jiabao 2 11 +Wen_Jiabao 4 7 +Wen_Jiabao 5 6 +Wen_Jiabao 7 9 +Wen_Jiabao 7 12 +Wen_Jiabao 8 12 +Wen_Jiabao 10 11 +Wen_Jiabao 11 12 +Wesley_Clark 1 2 +Yuri_Malenchenko 1 2 +Abbas_Kiarostami 1 Fujio_Mitarai 1 +Abdullah 1 Teresa_Heinz_Kerry 1 +Abdullah 3 Samuel_Waksal 1 +Abdullah 4 Julio_Cesar_Franco 1 +Abid_Hamid_Mahmud_Al-Tikriti 1 Anjum_Hussain 1 +Abid_Hamid_Mahmud_Al-Tikriti 2 Doris_Schroeder 1 +Adam_Freier 1 Hillary_Clinton 3 +Adam_Freier 1 Princess_Caroline 1 +Adam_Freier 1 Regina_Ip 1 +Alan_Ball 1 Kristin_Scott_Thomas 1 +Alan_Ball 2 Yuri_Malenchenko 1 +Albert_Costa 6 Gary_Barnett 1 +Albert_Costa 6 John_Marburger 1 +Albert_Costa 6 Wen_Jiabao 7 +Alberta_Lee 1 Babe_Ruth 1 +Alex_Popov 1 Kent_Rominger 2 +Alex_Popov 1 Matthew_During 1 +Ali_Khamenei 2 Roberto_Canessa 1 +Amelia_Vega 1 Gina_Lollobrigida 1 +Amelia_Vega 5 Jim_Harrick 1 +Amy_Pascal 1 Eduardo_Duhalde 7 +Amy_Pascal 1 Hank_Azaria 1 +Amy_Pascal 1 John_Marburger 1 +Amy_Redford 1 Roman_Coppola 1 +Amy_Redford 1 Victor_Kraatz 1 +Anderson_Varejao 1 Dennis_Oswald 1 +Anderson_Varejao 1 Garth_Drabinsky 1 +Andrei_Nikolishin 1 Angelica_Romero 1 +Andrew_Bernard 1 Don_Nickles 2 +Andrew_Bernard 1 Hassan_Wirajuda 1 +Andrew_Bernard 1 Mark_Broxmeyer 1 +Andrew_Fastow 1 Luca_Cordero_di_Montezemolo 1 +Andrew_Fastow 1 Meghann_Shaughnessy 1 +Andrew_Fastow 1 Tomas_Malik 1 +Andrew_Luster 1 Eric_Rosser 1 +Andrew_Luster 1 Jose_Manuel_Durao_Barroso 6 +Andrew_Luster 1 Lisa_Murkowski 1 +Anita_DeFrantz 1 Carla_Gay_Balingit 1 +Anita_DeFrantz 1 Penny_Lancaster 1 +Anjum_Hussain 1 David_Caraway 2 +Anne_Cavers 1 James_Barksdale 1 +Anne_Cavers 1 Stephen_Oake 1 +Anthony_Hazen 1 Debra_Yang 1 +Antonio_Catania 1 Barry_Bonds 1 +Antonio_Catania 1 Taylyn_Solomon 1 +Aretha_Franklin 1 Jean_Brumley 1 +Aretha_Franklin 1 Teruaki_Masumoto 1 +Art_Lopez 1 Diane_Lane 1 +Arthur_Johnson 1 Selma_Phoenix 1 +Arye_Mekel 2 Chung_Mong-joon 1 +Arye_Mekel 2 Georgina_Papin 1 +Babe_Ruth 1 Jim_Harrick 1 +Babe_Ruth 1 Pete_Aldridge 1 +Barry_Bonds 1 James_Barksdale 1 +Barry_Bonds 1 Mickey_Sherman 1 +Barry_Switzer 1 Jiang_Zemin 11 +Barry_Switzer 1 Kirk_Ferentz 2 +Bernard_Giraudeau 1 Keira_Knightley 1 +Bernard_Landry 3 Stacey_Jones 1 +Bernard_Landry 4 Sylvia_Plachy 1 +Biljana_Plavsic 1 Kieran_Prendergast 2 +Bill_Cartwright 1 Claude_Jorda 1 +Bill_Lerach 1 Martin_Verkerk 2 +Bob_Hope 1 Stanley_Ho 1 +Bob_Hope 5 Micah_Knorr 1 +Bob_Hope 5 Mitchell_Potter 1 +Bob_Menendez 1 David_Caraway 2 +Bob_Menendez 1 Lauren_Hutton 1 +Bob_Riley 1 Danny_Ainge 1 +Bob_Riley 1 Nastassia_Kinski 2 +Bob_Riley 1 Sheila_Wellstone 2 +Brad_Wilk 1 Phil_Cline 1 +Brad_Wilk 1 Scott_Dalton 1 +Brian_Grazier 1 Tomas_Malik 1 +Brian_StPierre 1 Steve_Mariucci 2 +Cari_Davis 1 Don_Henley 1 +Cari_Davis 1 Mark_Broxmeyer 1 +Carin_Koch 1 John_Banko 2 +Carla_Gay_Balingit 1 Frank_Lautenberg 2 +Carla_Gay_Balingit 1 Paula_Dobriansky 1 +Carlos_Ruiz 1 Peter_Schultz 1 +Carlos_Ruiz 3 Jennifer_Tilly 1 +Carly_Fiorina 2 Peter_Ueberroth 1 +Catherine_Woodard 1 Mike_Sweeney 1 +Chan_Ho_Park 1 Xiang_Xu 1 +Chance_Mock 1 Herb_Ritts 1 +Chance_Mock 1 Shafal_Mosed 1 +Chance_Mock 1 Simona_Hradil 1 +Chance_Mock 1 Troy_Jenkins 1 +Charles_Taylor 7 Tom_Curley 1 +Chea_Sophara 1 John_Wolf 1 +Chris_Cirino 1 Jen_Bice 1 +Chris_Cirino 1 Mary_Lou_Markakis 1 +Chris_Cirino 1 Peri_Gilpin 1 +Christina_Sawaya 1 Greg_Gilbert 1 +Chung_Mong-joon 1 David_Modell 1 +Chung_Mong-joon 1 Luca_Cordero_di_Montezemolo 1 +Chung_Mong-joon 1 Nick_Markakis 1 +Chung_Mong-joon 2 John_Rusnak 1 +Claire_Leger 1 David_Caraway 2 +Claire_Leger 1 Jerry_Sexton 1 +Claire_Leger 1 Nobuyuki_Idei 1 +Clark_Randt 1 Katie_Boone 1 +Clark_Randt 1 Marc-Andre_Fleury 1 +Clemente_de_la_Vega 1 Ron_Gonzales 1 +Colin_Campbell 1 Frank_Keating 1 +Colin_Campbell 1 Jean_Carnahan 2 +Colin_Campbell 1 Mitchell_Potter 1 +Colleen_Atwood 1 Penelope_Cruz 2 +Craig_Morgan 1 Matthew_McConaughey 1 +Dan_Duquette 1 Paddy_Torsney 1 +Daniel_Ortega 1 Norah_Jones 6 +Daniel_Patrick_Moynihan 1 Eglis_Yaima_Cruz 1 +Daniel_Patrick_Moynihan 1 Newton_Carlton_Slawson 1 +Daniel_Rouse 1 Doris_Schroeder 1 +Daniel_Rouse 1 Mike_Johanns 1 +Daniell_Sunjata 1 Marc-Andre_Fleury 2 +Danny_Ainge 1 Lee_Ann_Terlaji 1 +David_Leahy 1 Jerry_Sexton 1 +David_Scott_Morris 1 Valerie_Thwaites 1 +David_Zeplowitz 1 Jerry_Springer 1 +Dawn_Staley 1 Hasan_Wirayuda 1 +Debra_Rose 1 Jonathan_Edwards 2 +Debra_Yang 1 Hashim_Thaci 2 +Debra_Yang 1 Yekaterina_Guseva 1 +Denis_Fassou-Nguesso 1 William_Overlin 1 +Derrick_Taylor 1 Giuseppe_Morchio 1 +Derrick_Taylor 1 William_Overlin 1 +Dick_Cheney 5 James_Murdoch 1 +Dick_Cheney 8 Nobuyuki_Idei 1 +Dick_Smothers 1 Yoon_Won-Sik 1 +Dita_Von_Tesse 1 Mohammed_Al-Douri 12 +Don_Henley 1 Ernie_Eves 1 +Don_Henley 1 Gil_Cates 1 +Don_Henley 1 Stephen_Frears 1 +Don_Henley 1 Ziwang_Xu 1 +Donald_Regan 1 John_Gruden 1 +Doris_Schroeder 2 Joe_Mantegna 1 +Eduardo_Duhalde 1 Paula_Dobriansky 1 +Eduardo_Duhalde 2 Tab_Turner 1 +Eduardo_Duhalde 13 James_Murdoch 1 +Edward_Egan 1 Ernie_Fletcher 2 +Edward_Johnson 1 Jessica_Capshaw 1 +El_Hadji_Diouf 1 Hitomi_Soga 2 +Emilio_Botin 1 Luca_Cordero_di_Montezemolo 1 +Emilio_Botin 1 Robert_Hyatt 1 +Enrique_Iglesias 1 Gisele_Bundchen 2 +Eric_Bana 1 Mike_Sweeney 1 +Eric_Bana 1 Tab_Turner 1 +Eric_Benet 1 Mohammad_Mustapha_Miro 1 +Eric_Rosser 1 Marc-Andre_Fleury 2 +Eric_Rosser 1 Peter_Chan 1 +Ernie_Eves 2 Hana_Makhmalbaf 1 +Fernando_Vargas 1 Julio_Cesar_Franco 1 +Fernando_Vargas 3 Martin_Verkerk 3 +Flavia_Pennetta 1 Larry_Flynt 1 +Flavia_Pennetta 1 Micah_Knorr 1 +Flavia_Pennetta 1 Paul_Wellstone 1 +Frank_Keating 1 Paul_McCartney 7 +Frank_Lautenberg 2 Michael_Capellas 1 +Frank_Lautenberg 2 Steve_Allee 1 +Fujio_Mitarai 1 Harvey_Weinstein 1 +Gabriella_Bo 1 Jeremy_Shockey 2 +Gavyn_Arthur 1 Kevin_Hearn 1 +Gavyn_Arthur 1 Lynne_Slepian 1 +Gavyn_Arthur 1 Zahir_Shah 1 +George_HW_Bush 7 George_Maxwell_Richards 1 +George_HW_Bush 13 Penny_Lancaster 1 +George_Maxwell_Richards 1 Reggie_Miller 1 +Georgina_Papin 1 Mary_Lou_Retton 1 +Georgina_Papin 1 Svend_Robinson 1 +Gerry_Kelly 1 Keira_Knightley 1 +Giannina_Facio 1 Peri_Gilpin 1 +Gideon_Black 1 Kevin_Hearn 1 +Gideon_Black 1 Shafal_Mosed 1 +Gisele_Bundchen 2 Penny_Lancaster 1 +Giuseppe_Morchio 1 Penny_Lancaster 1 +Greg_Frers 1 Mohammed_Al-Douri 4 +Greg_Gilbert 1 Mehmet_Ali_Sahin 1 +Gregorio_Honasan 1 Jim_Schwarz 1 +Gregorio_Honasan 1 Jonathan_Fine 1 +Gregorio_Honasan 1 Matt_Doherty 1 +Hama_Arba_Diallo 1 Zhang_Yimou 1 +Hana_Makhmalbaf 1 Raquel_Welch 1 +Hassan_Wirajuda 1 Scott_Dalton 1 +Hector_Babenco 2 Talisa_Bratt 1 +Herb_Ritts 1 Mark_Broxmeyer 1 +Hugh_Campbell 1 Ken_Watanabe 1 +Hugh_Campbell 1 Masamori_Tokuyama 1 +Hugh_Campbell 1 Tom_Foy 1 +Imran_Khan 1 Matt_Morris 1 +Imran_Khan 1 Nancy_Humbert 1 +Ivan_Lee 1 Jennifer_Furminger 1 +James_Hallock 1 Robert_Beck 1 +James_Murdoch 1 Martina_McBride 3 +Jamie_Kellner 1 Tomas_Malik 1 +Jamie_King 1 Lauren_Hutton 2 +Jane_Fonda 2 Martha_Stewart 4 +Janet_Leigh 1 Marianne_Stanley 1 +Jaromir_Jagr 1 Troy_Jenkins 1 +Jean_Brumley 2 Jim_Harrick 1 +Jen_Bice 1 Judi_Dench 2 +Jennifer_Furminger 1 Matthew_During 1 +Jennifer_Tilly 1 Serge_Klarsfeld 1 +Jennifer_Tilly 1 Stella_Tennant 1 +Jeremy_Shockey 2 Robert_Beck 1 +Jerry_Regier 2 Paulina_Rodriguez_Davila 1 +Jerry_Regier 3 Lee_Hyung-taik 1 +Jerry_Regier 3 Pete_Rose 2 +Jessica_Capshaw 1 Penelope_Cruz 2 +Jiang_Zemin 4 Michelangelo_Antonioni 1 +Jiang_Zemin 13 Roman_Coppola 1 +Jim_Sterk 1 Jonathan_Edwards 3 +Jim_Sterk 1 Ken_Watanabe 2 +Joey_Harrington 1 Nick_Price 1 +John_Lynch 1 Michael_Sullivan 1 +John_Robbins 1 Tom_Foy 1 +John_Thune 1 Milan_Milutinovic 1 +John_Velazquez 1 Marco_Antonio_Barrera 3 +John_Velazquez 1 Tammy_Lynn_Michaels 1 +John_Wolf 2 Prince_Harry 1 +Jonathan_Byrd 1 Mike_Matheny 1 +Jonathan_Byrd 1 Sheikh_Ahmed_Yassin 1 +Jonathan_Horton 1 Lee_Ann_Terlaji 1 +Jorge_Alberto_Galindo 1 Roger_King 1 +Joseph_Biden 5 Mahdi_Al_Bassam 1 +Joseph_Biden 5 Tung_Chee-hwa 4 +Judy_Genshaft 1 Shane_Warne 1 +Jules_Asner 1 Robert_Beck 1 +Julien_Varlet 1 Scott_Dalton 1 +Kate_Burton 1 Martha_Stewart 2 +Kate_Richardson 1 Roman_Coppola 1 +Katie_Boone 1 Uthai_Pimchaichon 1 +Katie_Smith 1 Leland_Chapman 1 +Katrin_Susi 1 Luca_Cordero_di_Montezemolo 1 +Kay_Bailey_Hutchison 1 Theodore_Tweed_Roosevelt 3 +Kieran_Prendergast 1 Lachlan_Murdoch 1 +King_Abdullah_II 4 Robert_Hyatt 1 +Kirk_Ferentz 2 Milan_Milutinovic 1 +Lachlan_Murdoch 1 Mickey_Sherman 1 +Larry_Flynt 1 Phil_Cline 1 +Larry_Greene 1 Mike_Brey 2 +Laurence_Tribe 1 Uthai_Pimchaichon 1 +Lee_Ann_Terlaji 1 Roman_Coppola 1 +Lee_Yuan-tseh 1 Shafal_Mosed 1 +Lee_Yuan-tseh 1 Yuri_Malenchenko 2 +Leland_Chapman 1 Theodore_Tweed_Roosevelt 1 +Leslie_Moonves 2 Robert_Hyatt 1 +Lisa_Murkowski 1 Serge_Klarsfeld 1 +Magui_Serna 1 Maura_Tierney 1 +Mahdi_Al_Bassam 1 Marc-Andre_Fleury 2 +Mahdi_Al_Bassam 1 Mother_Teresa 1 +Malak_Habbak 1 Prince_Harry 3 +Manuel_Pellegrini 1 Rolf_Eckrodt 1 +Manuela_Montebrun 1 Martina_McBride 2 +Manuela_Montebrun 1 Steven_Seagal 2 +Marion_Fahnestock 1 Wen_Jiabao 3 +Mark_Hamister 1 Xiang_Xu 1 +Martha_Martinez_Flores 1 Mike_Brey 2 +Martha_Stewart 5 Victor_Kraatz 1 +Martin_Lawrence 1 Mickey_Sherman 1 +Martin_Lawrence 1 Penny_Lancaster 1 +Martin_McCauley 2 Steven_Seagal 1 +Martin_Verkerk 1 Steven_Seagal 1 +Martina_McBride 3 Peter_Schultz 1 +Masamori_Tokuyama 1 Regina_Ip 1 +Masamori_Tokuyama 1 Serge_Melac 1 +Matt_Doherty 2 Saoud_Al_Faisal 1 +Meghann_Shaughnessy 1 William_Perry 1 +Mehmet_Ali_Sahin 1 Peri_Gilpin 1 +Mehmet_Ali_Sahin 1 Raquel_Welch 2 +Michael_Capellas 1 Victor_Garber 1 +Mickey_Sherman 1 Roman_Coppola 1 +Mike_Brey 2 Nick_Markakis 1 +Mike_Johanns 1 Pascal_Lamy 1 +Mike_Matheny 1 Norm_Coleman 5 +Mohammad_Mustapha_Miro 1 Steve_Mariucci 2 +Mother_Teresa 1 Yekaterina_Guseva 1 +Nancy_Humbert 1 Park_Na-kyong 1 +Nanni_Moretti 1 Shigeru_Ishiba 1 +Natalia_Dmitrieva 1 Willie_Nelson 1 +Natalie_Juniardi 1 Uthai_Pimchaichon 1 +Nick_Price 1 Tab_Turner 1 +Norm_Coleman 4 Peri_Gilpin 1 +Norm_Coleman 7 Scott_McClellan 5 +Paddy_Torsney 1 Tung_Chee-hwa 9 +Park_Na-kyong 1 Vecdi_Gonul 1 +Patrick_Rafter 1 Peter_Care 1 +Pierre_Van_Hooijdonk 1 Scott_McClellan 1 +Regina_Ip 1 Rohman_al-Ghozi 1 +Renee_Zellweger 9 Yuri_Malenchenko 2 +Rolf_Eckrodt 2 Teresa_Heinz_Kerry 1 +Rubens_Barrichello 10 Stella_Tennant 1 +Saoud_Al_Faisal 1 Vecdi_Gonul 1 +Scott_Dalton 1 Zach_Safrin 1 +Scott_McClellan 2 Tom_Schnackenberg 1 +Todd_MacCulloch 1 Willie_Nelson 1 +Tom_Curley 1 Wanda_de_la_Jesus 1 +Troy_Jenkins 1 Walid_Al-Awadi 1 +William_Overlin 1 Yekaterina_Guseva 1 +Abdoulaye_Wade 1 2 +Abdoulaye_Wade 1 3 +Abdoulaye_Wade 2 3 +Adam_Sandler 1 2 +Adam_Sandler 1 4 +Adam_Sandler 2 3 +Aicha_El_Ouafi 1 3 +Aicha_El_Ouafi 2 3 +Akbar_Hashemi_Rafsanjani 1 3 +Akbar_Hashemi_Rafsanjani 2 3 +Al_Pacino 1 2 +Al_Pacino 1 3 +Alex_Barros 1 2 +Allyson_Felix 1 3 +Allyson_Felix 1 4 +Allyson_Felix 1 5 +Allyson_Felix 4 5 +Anastasia_Myskina 1 2 +Andy_Roddick 8 12 +Andy_Roddick 10 15 +Andy_Roddick 13 15 +Anna_Nicole_Smith 1 2 +Antonio_Palocci 1 8 +Antonio_Palocci 3 6 +Antonio_Palocci 4 5 +Antonio_Palocci 5 7 +Antonio_Palocci 6 8 +Arnoldo_Aleman 1 3 +Arnoldo_Aleman 3 5 +Ashton_Kutcher 1 3 +Ashton_Kutcher 2 3 +Augusto_Roa_Bastos 1 2 +Aung_San_Suu_Kyi 1 2 +Barry_Zito 1 2 +Bill_Graham 1 9 +Bill_Graham 3 4 +Bill_Graham 4 6 +Bill_Graham 5 6 +Bob_Dole 1 3 +Bruce_Weber 1 2 +Carlos_Mesa 1 2 +Carolyn_Dawn_Johnson 1 2 +Carolyn_Dawn_Johnson 2 3 +Celine_Dion 3 8 +Chakib_Khelil 1 2 +Chen_Shui-bian 2 4 +Chen_Shui-bian 3 5 +Christopher_Walken 1 3 +Christopher_Walken 1 4 +Claudia_Pechstein 1 2 +Claudia_Pechstein 1 4 +Claudia_Pechstein 3 4 +Claudia_Pechstein 3 5 +Claudia_Pechstein 4 5 +Clay_Aiken 2 4 +Clay_Aiken 3 4 +Clay_Aiken 3 5 +Colin_Powell 40 71 +Colin_Powell 49 234 +Colin_Powell 133 170 +Colin_Powell 182 198 +Cristina_Fernandez 1 2 +Daisy_Fuentes 2 3 +Damon_van_Dam 1 2 +Dan_Wheldon 1 2 +David_Coulthard 1 2 +David_Kelley 1 2 +Debra_Brown 1 2 +Dennis_Erickson 1 2 +Derek_Lowe 1 2 +Eddie_Sutton 1 2 +Edie_Falco 1 2 +Elijah_Wood 2 3 +Elizabeth_Hurley 1 4 +Elizabeth_Hurley 2 5 +Emily_Robison 1 2 +Ethan_Hawke 1 4 +Eunice_Barber 1 2 +Felix_Mantilla 1 2 +Fidel_Castro 1 18 +Fidel_Castro 3 7 +Fidel_Castro 5 8 +Fidel_Castro 8 12 +Fidel_Castro 11 13 +Francisco_Flores 1 2 +Francisco_Flores 1 3 +Frank_Dunham_Jr 1 2 +Franko_Simatovic 1 2 +Fred_Eckhard 1 2 +Fred_Eckhard 1 3 +Fred_Eckhard 2 3 +GL_Peiris 1 2 +GL_Peiris 1 3 +GL_Peiris 2 3 +GL_Peiris 2 4 +Garry_Kasparov 1 2 +Hassan_Nasrallah 1 2 +Heidi_Klum 1 3 +Heidi_Klum 1 4 +Heidi_Klum 2 4 +Heidi_Klum 3 4 +Heinz_Feldmann 1 2 +Heinz_Feldmann 2 3 +Iban_Mayo 1 2 +Imad_Moustapha 1 2 +Inam-ul-Haq 1 2 +James_Gandolfini 1 3 +James_Gandolfini 2 3 +Janet_Thorpe 1 2 +Jean-Pierre_Raffarin 1 2 +Jean-Pierre_Raffarin 1 6 +Jean-Pierre_Raffarin 3 4 +Jean-Pierre_Raffarin 3 5 +Jean-Pierre_Raffarin 4 7 +Jean-Pierre_Raffarin 5 7 +Jean-Pierre_Raffarin 6 7 +Jeffrey_Scott_Postell 1 2 +Jennifer_Capriati 2 14 +Jennifer_Capriati 7 32 +Jennifer_Capriati 33 42 +Job_Cohen 1 2 +John_McCormack 1 2 +John_Paul_II 1 4 +John_Paul_II 2 5 +John_Paul_II 2 8 +John_Paul_II 4 9 +John_Paul_II 10 11 +John_Ruiz 1 2 +John_Stallworth 1 2 +John_Stockton 2 4 +John_Travolta 2 6 +John_Travolta 3 5 +John_Travolta 5 7 +Jonathan_Mostow 1 2 +Jorge_Arce 1 2 +Joschka_Fischer 1 10 +Joschka_Fischer 6 11 +Joschka_Fischer 7 11 +Joschka_Fischer 11 17 +Joschka_Fischer 15 16 +Jose_Canseco 1 3 +Juan_Manuel_Marquez 1 2 +Juan_Manuel_Marquez 1 3 +Juan_Manuel_Marquez 2 3 +Juan_Valencia_Osorio 1 2 +Julie_Gerberding 9 13 +Julie_Gerberding 12 15 +Kate_Hudson 1 4 +Kate_Hudson 1 8 +Kate_Hudson 2 3 +Kate_Hudson 4 9 +Kate_Hudson 6 7 +Kemal_Dervis 1 3 +Kemal_Dervis 2 3 +Kenneth_Evans 1 2 +Kifah_Ajouri 1 2 +Larry_Lucchino 1 2 +Latrell_Sprewell 1 2 +Lech_Walesa 1 2 +Lee_Tae-sik 1 2 +Lisa_Marie_Presley 1 3 +Liza_Minnelli 2 3 +Liza_Minnelli 3 4 +Liza_Minnelli 3 6 +Liza_Minnelli 4 5 +Liza_Minnelli 5 6 +Liza_Minnelli 6 7 +Madonna 1 4 +Madonna 2 3 +Madonna 4 5 +Mariah_Carey 3 6 +Mary_Tyler_Moore 1 2 +Mathias_Reichhold 1 2 +Matt_Damon 1 2 +Matt_Damon 1 3 +Matt_Damon 2 4 +Matt_Damon 3 4 +Maureen_Fanning 1 2 +Melanie_Griffith 1 2 +Melanie_Griffith 1 3 +Melanie_Griffith 2 3 +Michael_Ballack 1 2 +Michael_Winterbottom 1 3 +Michael_Winterbottom 2 3 +Michelle_Collins 1 2 +Milo_Maestrecampo 1 2 +Mohamed_Benaissa 1 2 +Mohamed_ElBaradei 2 4 +Mohamed_ElBaradei 3 8 +Morgan_Freeman 1 2 +Muhammad_Ali 1 3 +Muhammad_Ali 1 7 +Muhammad_Ali 2 5 +Muhammad_Ali 6 10 +Muhammad_Ali 7 9 +Mukesh_Ambani 1 2 +Mukesh_Ambani 1 3 +Paris_Hilton 1 2 +Pat_Cox 1 2 +Paul_Burrell 3 7 +Paul_Burrell 5 11 +Paul_Burrell 8 10 +Paul_Wolfowitz 8 10 +Paula_Radcliffe 1 2 +Paula_Radcliffe 1 3 +Paula_Radcliffe 2 3 +Paula_Radcliffe 2 4 +Paula_Radcliffe 3 4 +Paula_Radcliffe 3 5 +Paula_Radcliffe 4 5 +Paulo_Cesar_Pinheiro 1 2 +Pedro_Solbes 1 3 +Pedro_Solbes 1 4 +Pedro_Solbes 2 3 +Pedro_Solbes 3 4 +Pete_Carroll 1 2 +Pete_Carroll 1 3 +Pete_Carroll 2 3 +Pete_Sampras 2 12 +Pete_Sampras 2 13 +Pete_Sampras 3 15 +Pete_Sampras 4 20 +Pete_Sampras 6 7 +Pete_Sampras 6 8 +Pete_Sampras 10 13 +Pete_Sampras 12 15 +Peter_Struck 1 5 +Peter_Struck 2 5 +Phil_Vassar 1 2 +Pierre_Boulanger 1 2 +Prince_Willem-Alexander 1 2 +Prince_Willem-Alexander 1 3 +Prince_Willem-Alexander 2 3 +Queen_Elizabeth_II 3 7 +Queen_Elizabeth_II 9 12 +Queen_Elizabeth_II 10 11 +Queen_Elizabeth_II 10 12 +Ray_Nagin 1 2 +Ricardo_Maduro 1 2 +Richard_Branson 1 2 +Richard_Virenque 1 4 +Richard_Virenque 1 6 +Richard_Virenque 1 7 +Richard_Virenque 2 7 +Richard_Virenque 2 8 +Rick_Carlisle 2 4 +Rick_Wagoner 1 2 +Robbie_Williams 1 2 +Robbie_Williams 1 3 +Roberto_Carlos 2 3 +Roberto_Carlos 2 4 +Roseanne_Barr 1 2 +Roseanne_Barr 2 3 +Ruben_Studdard 1 2 +Sammy_Sosa 1 2 +Sarah_Jessica_Parker 1 3 +Sarah_Jessica_Parker 2 4 +Sarah_Jessica_Parker 3 4 +Sharon_Davis 1 2 +Shaul_Mofaz 1 2 +Shaul_Mofaz 2 3 +Stan_Heath 1 2 +Svetlana_Koroleva 1 2 +Terrell_Suggs 1 2 +Tim_Henman 2 12 +Tim_Henman 8 19 +Tom_Daschle 7 8 +Tom_Daschle 15 21 +Tom_Daschle 15 22 +Tony_Curtis 1 2 +Valentino_Rossi 1 2 +Valentino_Rossi 2 4 +Valentino_Rossi 3 6 +Valentino_Rossi 4 5 +Valentino_Rossi 5 6 +Vanessa_Redgrave 1 3 +Vanessa_Redgrave 1 4 +Vanessa_Redgrave 2 5 +Vanessa_Redgrave 3 4 +Victoria_Clarke 1 5 +Vladimiro_Montesinos 1 2 +Vladimiro_Montesinos 1 3 +Vladimiro_Montesinos 2 3 +Wayne_Ferreira 1 2 +Wayne_Ferreira 1 3 +Wayne_Ferreira 1 5 +Wayne_Ferreira 2 5 +Wayne_Ferreira 3 4 +Will_Smith 1 2 +Yasser_Arafat 1 6 +Yasser_Arafat 1 8 +Yasser_Arafat 2 3 +Yasser_Arafat 2 5 +Yasser_Arafat 3 4 +Yasser_Arafat 3 8 +Yasser_Arafat 4 5 +Yasser_Arafat 5 8 +Yuri_Fedotov 1 2 +Zoran_Djindjic 1 3 +Zoran_Djindjic 1 4 +Aaron_Patterson 1 Frank_Bell 1 +Abdoulaye_Wade 4 Bruce_Weber 2 +Abner_Martinez 1 Carlos_Alberto 1 +Adam_Sandler 2 Matthew_Ouimet 1 +Adam_Sandler 3 Saeed_Anwar 1 +Adolfo_Aguilar_Zinser 3 Jaime_Pressly 1 +Agnelo_Queiroz 1 Aung_San_Suu_Kyi 2 +Agnelo_Queiroz 1 Dave_Barr 1 +Aicha_El_Ouafi 3 Michael_Lechner 1 +Akbar_Hashemi_Rafsanjani 1 Larry_Harris 1 +Al_Pacino 2 Charles_Cope 1 +Alex_Barros 1 Brandon_Jones 1 +Alex_Barros 2 Will_Smith 2 +Alex_Ferguson 1 Rainer_Gut 1 +Alex_Wallau 1 Shireen_Amir_Begum 1 +Alexandra_Jackson 1 Larry_Harris 1 +Alfonso_Portillo 1 Benito_Santiago 1 +Alfonso_Portillo 1 Faye_Alibocus 1 +Alfonso_Portillo 1 Fidel_Castro 17 +Ali_Abdullah_Saleh 1 Khalid_Qazi 1 +Allan_Houston 1 Andy_Garcia 1 +Allan_Houston 1 Heidi_Klum 1 +Allan_Houston 1 Thomas_Mesereau_Jr 1 +Ally_Sheedy 1 Hugh_Carey 1 +Ally_Sheedy 1 Myung_Yang 1 +Amanda_Marsh 1 Tony_Curtis 2 +Anastasia_Myskina 1 Raul_Gonzalez 1 +Anastasia_Myskina 3 Len_Jenoff 2 +Andrzej_Tyszkiewicz 1 Wes_Craven 1 +Andy_Griggs 1 Lech_Walesa 2 +Andy_Rooney 1 Jessica_Simpson 1 +Anna_Nicole_Smith 2 Marcus_Garrettson 1 +Antonio_Palocci 3 Liza_Minnelli 1 +Antonio_Palocci 5 JC_Chasez 1 +Antonio_Palocci 5 Jose_Woldenberg 1 +Antonio_Palocci 6 John_Geoghan 1 +Antonio_Palocci 8 Hans_Corell 1 +Arif_Mardin 1 Eduardo_Fischer 1 +Arnaud_Lagardere 1 Melanie_Griffith 3 +Ashton_Kutcher 2 Daniel_Barenboim 1 +Asif_Hanif 1 Robbie_Williams 1 +Asmaa_Assad 1 Barry_Hinson 1 +Aung_San_Suu_Kyi 1 Charla_Moye 1 +Azmi_Bishara 1 Sammy_Sosa 2 +Barry_Hinson 1 Nino_DAngelo 1 +Barry_Zito 2 Chris_Gratton 1 +Bill_Graham 8 Michelle_Hofland 1 +Bill_Graham 9 Jacqueline_Marris 1 +Bill_Readdy 1 Brendan_Gaughan 1 +Bill_Readdy 1 Jaymon_Crabb 1 +Bill_Readdy 1 Yasser_Arafat 3 +Billy_Rork 1 Eva_Mendes 1 +Billy_Rork 1 German_Khan 1 +Billy_Rork 1 Peter_Struck 2 +Bison_Dele 1 Brian_McIntyre 1 +Bob_Dole 2 Dai_Chul_Chyung 1 +Bob_Dole 2 John_Henry 1 +Bob_Dole 3 Chris_Gratton 1 +Bob_Dole 3 Hugh_Carey 1 +Bob_Geldof 1 Zoran_Djindjic 2 +Bob_Geldof 2 Ed_Wade 1 +Bob_Holden 1 Fernando_Leon_de_Aranoa 1 +Bob_Iger 1 Edie_Falco 1 +Bob_Iger 1 Jean-Claude_Van_Damme 1 +Brian_Clemens 1 Brian_Meadors 1 +Brian_Clemens 1 Melanie_Griffith 2 +Brian_Heidik 2 Djabir_Said-Guerni 1 +Brian_McIntyre 1 Hans_Corell 1 +Brian_McIntyre 1 Mohammed_Abulhasan 1 +Brian_Pavlich 1 Ruben_Wolkowyski 1 +Brook_Robinson 1 Tom_McClintock 1 +Brooke_Adams 1 Paula_Prentiss 1 +Brooke_Gordon 1 Joschka_Fischer 12 +Bruce_Weber 1 Hal_Sellers 1 +Bryan_Thomas 1 Joey_Mantia 1 +Bustam_A_Zedan_Aljanabi 1 Kajsa_Bergqvist 1 +Calvin_Joseph_Coleman 1 Hassan_Nasrallah 1 +Carla_Sullivan 1 Edie_Falco 1 +Carlos_Barragan 1 Chen_Shui-bian 3 +Carlos_Salinas 1 Norman_Mailer 1 +Carlos_Salinas 1 Sonya_Walger 1 +Carolyn_Dawn_Johnson 3 Lydia_Shum 1 +Carolyn_Kuhl 1 Pierre_Boulanger 1 +Celine_Dion 2 John_Stallworth 2 +Celine_Dion 5 Linda_Baboolal 1 +Celine_Dion 5 Tom_Poston 1 +Chakib_Khelil 2 Chuck_Woolery 1 +Charla_Moye 1 Patti_Balgojevich 1 +Charles_Cope 1 Garry_Alejano 1 +Charles_Holzner 1 Eurico_Guterres 1 +Charles_Holzner 1 Greg_Kinnear 1 +Chen_Shui-bian 3 Gaston_Gaudio 1 +Chris_Gratton 1 Mario_Vasquez_Rana 1 +Chris_Kolanas 1 Joshua_Gracin 1 +Claudia_Pechstein 2 Mireille_Jospin-Dandieu 1 +Clay_Aiken 1 Svetlana_Koroleva 1 +Colin_Powell 95 Frank_Hsieh 1 +Craig_David 1 Tom_McClintock 1 +Craig_Wilson 1 Kajsa_Bergqvist 1 +Cristina_Fernandez 2 Stephen_Cooper 1 +Curtis_Joseph 1 Terrell_Suggs 2 +Cynthia_Rowley 1 Michael_Friedman 1 +Damon_van_Dam 2 Jason_Sorens 1 +Daniel_Barenboim 1 Tyler_Grillo 1 +Daniel_Bruehl 1 Gus_Frerotte 1 +Daniel_Bruehl 1 Max_Mosley 1 +Daniel_Bruehl 1 Ramon_Cardenas 1 +Daniele_Bergamin 1 Kenneth_Evans 1 +Danielle_Spencer 1 Rachel_Wheatley 1 +Darcy_Regier 1 William_Hurt 1 +Dave_Matthews 1 Linda_Dano 1 +Dave_Matthews 1 Paul_Burrell 7 +Dave_McNealey 1 Gerald_Riley 1 +David_Bisbal 1 Terri_Clark 1 +David_Chase 1 Ekaterina_Dmitriev 1 +David_McCullough 1 Evo_Morales 1 +David_McKiernan 1 Fernando_Leon_de_Aranoa 1 +David_McKiernan 1 Jane_Leeves 1 +Dennis_Erickson 2 Hisham_Halawi 1 +Dennis_Erickson 2 Noer_Muis 1 +Derek_Lowe 2 Hisham_Halawi 1 +Derek_Lowe 2 Madonna 1 +Derek_Lowe 2 Mohamed_Benaissa 1 +Diego_Colorado 1 Jason_Sorens 1 +Diego_Colorado 1 John_Gordnick 1 +Diego_Diego_Lerman 1 Francisco_Flores 2 +Diego_Diego_Lerman 1 Lesley_Coppin 1 +Djabir_Said-Guerni 1 Paul_Tracy 1 +Djabir_Said-Guerni 1 Pedro_Solbes 3 +Dominik_Hrbaty 1 Toshi_Izawa 1 +Donald_Keck 1 Kyoko_Nakayama 1 +Donald_Keck 1 Tom_Poston 1 +Donna_Walker 1 Jacqueline_Marris 1 +Dragan_Covic 1 Todd_Reid 1 +Dudley_Rogers 1 Syed_Ibrahim 1 +Dunn_Lampton 1 Jessica_Simpson 1 +Dustin_Hoffman 1 Nicole_Parker 1 +Ed_Wade 1 Roger_Staubach 1 +Ed_Wade 1 Terri_Clark 1 +Eddie_Lucio 1 Patti_Balgojevich 1 +Eddie_Sutton 1 James_Wattana 1 +Eddie_Sutton 1 Jeanne_Anne_Schroeder 1 +Eddie_Sutton 2 Ronald_Ito 1 +Eduardo_Fischer 1 Kimberly_Bruckner 1 +Edward_Lohn 1 Lily_Safra 1 +Edward_Lohn 1 Nino_DAngelo 1 +Ekaterina_Dmitriev 1 Mitch_Kupchak 1 +Eladio_Larez 1 Frank_Pallone 1 +Eli_Broad 1 Ravan_AG_Farhadi 1 +Elijah_Wood 2 Stefano_Gabbana 1 +Elijah_Wood 3 Marcus_Garrettson 1 +Elijan_Ingram 1 Michelle_Hofland 1 +Elijan_Ingram 1 Nastia_Liukin 1 +Elvis_Costello 1 Jaime_Pressly 1 +Emelie_Loit 1 Thomas_Mesereau_Jr 1 +Eric_Staal 1 Jerry_Lewis 1 +Erin_Hershey_Presley 1 Frank_Dunham_Jr 1 +Erin_Hershey_Presley 1 Yukio_Hatoyama 1 +Eva_Mendes 1 Larry_Harris 1 +Faye_Alibocus 1 Frank_Bell 1 +Faye_Alibocus 1 Tommy_Tubberville 1 +Felix_Mantilla 1 Jerry_Lewis 1 +Fernando_Leon_de_Aranoa 1 John_Scarlett 1 +Fernando_Leon_de_Aranoa 1 Mike_Leach 1 +Fidel_Castro 7 Hu_Maoyuan 1 +Francisco_Flores 4 Hisham_Halawi 1 +Frank_Bell 1 Khatol_Mohammad_Zai 1 +Frank_Dunham_Jr 1 Tom_McClintock 1 +Frank_Hsieh 1 Paula_Prentiss 1 +Frank_Pallone 1 Jim_Wall 1 +Frank_Pallone 1 Mary_Tyler_Moore 2 +Frank_Schmoekel 1 Ronald_Ito 1 +Franko_Simatovic 2 Tyler_Grillo 1 +Franz_Gsell 1 John_Scarlett 1 +Franz_Gsell 1 Sarah_Jessica_Parker 2 +Fred_Eckhard 1 Rachel_Wheatley 1 +Garry_Alejano 1 Iban_Mayo 1 +Garry_Alejano 1 Michael_Olowokandi 1 +Garry_Alejano 1 Morgan_Freeman 2 +Garry_Alejano 1 Peter_Struck 5 +Gary_Marshall 1 Rashid_Qureshi 1 +Gerald_Fitch 1 Robin_Williams 1 +Gerald_Riley 1 Michael_Hagee 1 +German_Khan 1 Morgan_Freeman 2 +Grady_Little 1 Robert_Morvillo 1 +Hal_Sellers 1 Janet_Thorpe 1 +Hans_Corell 1 John_Gordnick 1 +Heidi_Klum 2 Paul_Reiser 1 +Heidi_Klum 3 Pedro_Solbes 1 +Hermando_Harton 1 Paris_Hilton 2 +Holly_Robinson_Peete 1 Michael_Hagee 1 +Howard_Ross 1 Kajsa_Bergqvist 1 +Howard_Ross 1 Patsy_Hardy 1 +Hugh_Carey 1 Lawrence_Roberts 1 +Hugh_Carey 1 Tracee_Treadwell 1 +Iban_Mayo 2 Juan_Valencia_Osorio 1 +Iban_Mayo 2 Rosalie_Perkov 1 +Ibrahim_Al-Marashi 1 John_Travolta 3 +Ibrahim_Al-Marashi 1 Joshua_Gracin 1 +Imad_Moustapha 2 Kyoko_Nakayama 1 +Ira_Einhorn 1 John_Ruiz 2 +Jacqueline_Marris 1 Matthew_Ouimet 1 +Jaime_Pressly 1 Tiago_Splitter 1 +James_Brown 1 Mary_Blige 1 +James_Brown 1 Paul_Reiser 1 +James_Mathis 1 Yuri_Fedotov 1 +James_Wattana 1 Paul_Reiser 1 +James_Wattana 1 Ron_Kirk 1 +James_Young 1 Jeffrey_Scott_Postell 2 +James_Young 1 Ronald_Harwood 1 +Jane_Leeves 1 Tom_McClintock 1 +Janet_Crawford 1 Thomas_Scavone 1 +Jason_Sorens 1 Lidija_Djukanovic 1 +Jason_Vale 1 Marcus_Garrettson 1 +Jason_Vale 1 Zoran_Djindjic 3 +Jaymon_Crabb 1 Ted_Christopher 1 +Jeff_Weaver 1 Kyoko_Nakayama 1 +Jeff_Weaver 1 Rosalie_Perkov 1 +Jeffrey_Scott_Postell 1 Tara_Dawn_Christensen 1 +Jeffrey_Scott_Postell 2 Lesley_Coppin 1 +Jennifer_Capriati 26 Ruben_Wolkowyski 1 +Jim_Bunning 1 Terence_Newman 1 +Jim_Jeffords 1 Peter_Rasch 1 +Jim_Jeffords 1 Tom_Sizemore 1 +Job_Cohen 1 John_Stockton 2 +John_Franco 1 Paul_Wolfowitz 4 +John_Gordnick 1 Linda_Dano 1 +John_Gordnick 1 Paul_Vathis 1 +John_Gordnick 1 Yasser_Arafat 3 +John_Stockton 4 Patsy_Hardy 1 +John_Travolta 2 Lech_Walesa 1 +Juan_Valencia_Osorio 1 Tom_Daschle 6 +Juergen_Schrempp 1 Mitt_Romney 1 +Julien_Boutter 1 Saeed_Anwar 1 +Kaisser_Bazan 1 Lawrence_Vito 1 +Kaisser_Bazan 1 Pierre_Lacroix 1 +Kajsa_Bergqvist 1 Tayshaun_Prince 1 +Keith_Snyder 1 Terri_Clark 1 +Kemal_Dervis 3 Roger_Staubach 1 +Kifah_Ajouri 1 Laurie_Pirtle 1 +Kifah_Ajouri 1 Raul_Gonzalez 1 +Kimberly_Bruckner 1 Pete_Carroll 1 +Kirk_Douglas 1 Nida_Blanca 1 +Larry_Harris 1 Michael_Lechner 1 +Larry_Lucchino 1 Rudy_Tomjanovich 1 +Latrell_Sprewell 1 Rosalie_Perkov 1 +Laura_Schlessinger 1 Tony_Curtis 1 +Laurie_Pirtle 1 Roberta_Combs 1 +Lawrence_Roberts 1 Lynne_Thigpen 1 +Lawrence_Roberts 1 Nick_Cassavetes 1 +Lech_Walesa 2 Rick_Wagoner 2 +Len_Jenoff 2 Tom_Daschle 7 +Lew_Rywin 1 Terri_Clark 1 +Linda_Dano 1 Terence_Newman 1 +Lydia_Shum 1 Mario_Vasquez_Rana 1 +Lynne_Thigpen 1 Mary_Blige 1 +Malcolm_Glazer 1 Noer_Muis 1 +Malcolm_Glazer 1 Tabare_Vazquez 1 +Marc_Bulger 1 Paul_Wolfowitz 7 +Marcus_Garrettson 1 Pham_Sy_Chien 1 +Marcus_Garrettson 1 Roberto_Carlos 3 +Maribel_Dominguez 1 Michael_Winterbottom 3 +Marion_Barry 1 Steve_Peace 1 +Mark_Hogan 1 Queen_Elizabeth_II 11 +Mark_Hogan 1 Roger_Toussaint 1 +Mary_Blige 1 Raul_Gonzalez 1 +Massoud_Barzani 1 Pierre_Lacroix 1 +Max_Mosley 1 Raul_Gonzalez 1 +Melvin_Talbert 1 Rudy_Tomjanovich 1 +Michael_Hagee 1 Sonya_Walger 1 +Michael_Killeen 1 Sammy_Sosa 2 +Michael_Olowokandi 1 Pete_Carroll 2 +Michael_Winterbottom 1 Pierre_Boulanger 1 +Mike_Price 2 Tom_Miller 1 +Milo_Maestrecampo 1 Yuri_Fedotov 1 +Milo_Maestrecampo 2 Steve_Peace 1 +Milo_Maestrecampo 3 Stacey_Dales-Schuman 1 +Mitt_Romney 1 Nida_Blanca 1 +Mohamed_ElBaradei 8 Rachel_Wheatley 1 +Mohammad_Aktar 1 Rafeeuddin_Ahmed 1 +Mohammad_Aktar 1 Tom_McClintock 1 +Mohammed_Abulhasan 1 Stephane_Rousseau 1 +Mohammed_Abulhasan 1 Vladimiro_Montesinos 1 +Nastia_Liukin 1 Nicole_Parker 1 +Nastia_Liukin 1 Sharon_Davis 2 +Nicolas_Latorre 1 Paula_Prentiss 1 +Nicolas_Latorre 1 Vassilis_Xiros 1 +Nicolas_Latorre 1 William_Delahunt 1 +Pete_Sampras 10 Stan_Heath 1 +Phil_Vassar 2 Tabare_Vazquez 1 +Philip_Zalewski 1 Stefan_Holm 1 +Roberto_Carlos 1 Yasser_Arafat 5 +Roger_Cook 1 Wilbert_Foy 1 +Roger_Staubach 1 Severino_Antinori 1 +Ron_Kirk 1 Troy_Hudson 1 +Sammy_Sosa 2 Stan_Heath 2 +Sandra_Ceccarelli 1 Stephen_Cooper 1 +Sandra_Ceccarelli 1 Tom_Coughlin 1 +Shireen_Amir_Begum 1 Sushma_Swaraj 1 +Sinead_OConnor 1 Stephane_Rochon 1 +Aitor_Gonzalez 1 2 +Alec_Baldwin 2 4 +Allison_Janney 1 2 +Alvaro_Noboa 1 3 +Alvaro_Noboa 2 3 +Amanda_Coetzer 1 2 +Amer_al-Saadi 1 3 +Amer_al-Saadi 1 4 +Amer_al-Saadi 2 4 +Ana_Guevara 2 7 +Ana_Guevara 3 7 +Anneli_Jaatteenmaki 1 2 +Ari_Fleischer 6 9 +Ari_Fleischer 7 12 +Arianna_Huffington 1 2 +Arianna_Huffington 1 4 +Arianna_Huffington 2 3 +Arianna_Huffington 2 4 +Arnaud_Clement 1 2 +Arsinee_Khanjian 1 2 +Art_Howe 1 2 +Art_Howe 1 3 +Art_Howe 1 4 +Art_Howe 2 3 +Art_Howe 3 4 +Ben_Affleck 2 3 +Ben_Affleck 2 4 +Ben_Affleck 2 7 +Ben_Affleck 3 6 +Ben_Affleck 5 6 +Betsy_Smith 1 2 +Bill_Callahan 1 3 +Blythe_Hartley 1 2 +Bob_Huggins 1 3 +Bob_Huggins 3 4 +Bobby_Goldwater 1 2 +Bono 1 2 +Bono 1 3 +Bono 2 3 +Brad_Garrett 1 3 +Brad_Garrett 2 3 +Brad_Garrett 2 4 +Brian_Mulroney 1 2 +Bud_Selig 1 2 +Bud_Selig 2 3 +Bud_Selig 2 4 +Carla_Del_Ponte 1 2 +Carla_Del_Ponte 2 4 +Carla_Del_Ponte 2 5 +Carla_Del_Ponte 3 5 +Carlos_Ghosn 1 2 +Carlos_Manuel_Pruneda 1 2 +Carlos_Manuel_Pruneda 1 3 +Carlos_Menem 2 21 +Carlos_Menem 5 18 +Carlos_Menem 8 15 +Carlos_Menem 11 12 +Carlos_Menem 14 21 +Carlos_Menem 16 18 +Carmen_Electra 3 4 +Carmen_Electra 4 6 +Charlotte_Rampling 1 2 +Chick_Hearn 1 2 +Christine_Todd_Whitman 2 3 +Christine_Todd_Whitman 5 6 +Christopher_Reeve 1 3 +Chuck_Amato 1 2 +Cindy_Crawford 2 3 +Cindy_Margolis 1 2 +Claire_Danes 1 3 +Claire_Danes 2 3 +Conan_OBrien 1 2 +Conan_OBrien 2 3 +Conan_OBrien 2 4 +Conchita_Martinez 1 2 +Conchita_Martinez 1 3 +Dan_Morales 1 2 +Dan_Morales 1 3 +David_Hyde_Pierce 1 2 +David_Hyde_Pierce 1 3 +David_Hyde_Pierce 2 3 +David_Hyde_Pierce 2 4 +David_Myers 1 2 +David_Nalbandian 1 3 +David_Nalbandian 2 9 +David_Nalbandian 3 4 +David_Nalbandian 4 13 +David_Nalbandian 11 12 +David_Stern 1 2 +David_Stern 2 3 +Derek_Jeter 2 3 +Derek_Jeter 2 4 +Donatella_Versace 1 3 +Donatella_Versace 2 3 +Donna_Shalala 1 2 +Edmund_Hillary 1 2 +Edmund_Hillary 1 3 +Edmund_Hillary 2 3 +Elisabeth_Schumacher 1 2 +Elizabeth_Smart 3 5 +Erin_Runnion 1 3 +Erin_Runnion 2 3 +Erin_Runnion 2 4 +Erin_Runnion 3 4 +Fernando_Henrique_Cardoso 1 2 +Fernando_Henrique_Cardoso 1 3 +Fernando_Henrique_Cardoso 1 4 +Fernando_Henrique_Cardoso 2 7 +Fernando_Henrique_Cardoso 5 7 +Fernando_Henrique_Cardoso 6 7 +Francis_Mer 1 2 +Franz_Fischler 1 2 +Gary_Locke 1 2 +Gerry_Adams 1 3 +Gerry_Adams 1 7 +Gerry_Adams 2 6 +Gerry_Adams 3 5 +Gerry_Adams 4 6 +Gerry_Adams 5 6 +Gianna_Angelopoulos-Daskalaki 1 2 +Gil_de_Ferran 1 5 +Gil_de_Ferran 2 4 +Gil_de_Ferran 3 5 +Goh_Kun 1 2 +Grady_Irvin_Jr 1 2 +Gunter_Pleuger 3 5 +Gunter_Pleuger 3 6 +Harry_Belafonte 1 2 +Harry_Schmidt 1 3 +Harry_Schmidt 1 4 +Harry_Schmidt 2 3 +Harry_Schmidt 3 4 +Helen_Clark 2 4 +Hermann_Maier 1 2 +Ian_Thorpe 1 7 +Ian_Thorpe 1 10 +Ian_Thorpe 3 9 +Ian_Thorpe 5 7 +Ian_Thorpe 6 10 +Isabelle_Huppert 1 2 +James_Butts 1 2 +Jan-Michael_Gambill 1 2 +Jan-Michael_Gambill 2 3 +Jean-Francois_Pontal 1 3 +Jean-Marc_de_La_Sabliere 1 2 +Jean-Sebastien_Giguere 1 2 +Jeb_Bush 1 6 +Jeb_Bush 2 6 +Jeb_Bush 5 7 +Jeffrey_Jones 1 2 +Jim_OBrien 1 2 +Jim_OBrien 1 3 +Joe_Gatti 1 2 +John_Cusack 1 2 +John_Edwards 1 3 +John_Edwards 1 7 +John_Edwards 2 5 +John_Edwards 3 7 +John_Edwards 4 6 +John_Edwards 4 8 +John_Edwards 6 7 +John_Spencer 1 2 +John_Taylor 1 2 +John_Walsh 1 2 +Jolanta_Kwasniewski 1 2 +Jose_Mourinho 1 2 +Juergen_Peters 1 2 +Kalpana_Chawla 1 4 +Kalpana_Chawla 2 3 +Kalpana_Chawla 2 5 +Kalpana_Chawla 3 4 +Kelli_White 1 2 +Kim_Jin-sun 1 2 +Lana_Clarkson 1 2 +Laura_Hernandez 1 2 +Laura_Linney 1 2 +Laura_Linney 1 3 +Laura_Linney 2 4 +Laura_Linney 3 4 +Lenny_Wilkens 1 2 +Lenny_Wilkens 1 3 +Lenny_Wilkens 2 3 +Lloyd_Ward 1 2 +Mahathir_Mohamad 1 7 +Mahathir_Mohamad 1 12 +Mahathir_Mohamad 6 10 +Mark_Cuban 1 2 +Mark_Heller 1 2 +Mark_Schweiker 1 2 +Marty_Mornhinweg 1 2 +Marty_Mornhinweg 1 3 +Marty_Mornhinweg 2 3 +Michael_Phelps 1 2 +Michael_Phelps 2 4 +Mick_Jagger 2 4 +Miguel_Estrada 1 2 +Mike_Myers 1 6 +Mike_Myers 2 3 +Mike_Myers 2 4 +Mike_Myers 2 5 +Nadine_Vinzens 1 2 +Nathan_Lane 1 2 +Nicolas_Cage 1 2 +Nicolas_Cage 2 4 +Nicolas_Cage 3 4 +Nicole_Kidman 3 11 +Nicole_Kidman 8 17 +Nicole_Kidman 11 13 +Nicole_Kidman 13 14 +Nicole_Kidman 15 19 +Omar_Sharif 1 2 +Omar_Sharif 1 3 +Omar_Sharif 1 4 +Omar_Sharif 2 3 +Parris_Glendening 1 2 +Patrick_Leahy 1 2 +Patrick_McEnroe 1 2 +Patrick_Stewart 1 2 +Paul_Gascoigne 1 3 +Peter_Hillary 1 2 +Phan_Van_Khai 1 2 +Phan_Van_Khai 1 3 +Phan_Van_Khai 2 3 +Prince_Claus 1 4 +Prince_Claus 2 3 +Prince_Claus 2 4 +Raghad_Saddam_Hussein 1 2 +Ray_Allen 1 2 +Ray_Allen 2 3 +Richard_Crenna 1 2 +Richard_Gere 1 10 +Richard_Gere 6 9 +Richard_Myers 1 2 +Richard_Myers 1 4 +Richard_Myers 4 8 +Richard_Myers 15 16 +Rick_Romley 1 3 +Rob_Lowe 1 3 +Rob_Lowe 1 4 +Rob_Lowe 2 3 +Rob_Lowe 3 4 +Robby_Ginepri 1 2 +Robert_Witt 1 2 +Rod_Blagojevich 1 2 +Rolandas_Paksas 1 2 +Russell_Coutts 1 2 +Ruth_Dreifuss 1 2 +Sally_Kirkland 1 4 +Sean_Patrick_OMalley 1 3 +Sebastian_Saja 1 2 +Sebastian_Saja 1 3 +Sila_Calderon 1 2 +Sila_Calderon 1 3 +Sila_Calderon 1 4 +Sila_Calderon 2 3 +Stanley_Tong 1 2 +Steve_Park 1 2 +Susie_Castillo 1 2 +Sylvester_Stallone 1 5 +Sylvester_Stallone 2 4 +Sylvester_Stallone 3 8 +Sylvester_Stallone 4 5 +Takashi_Sorimachi 1 2 +Tang_Jiaxuan 2 6 +Tang_Jiaxuan 2 11 +Tang_Jiaxuan 4 9 +Tang_Jiaxuan 5 9 +Tang_Jiaxuan 7 10 +Terje_Roed-Larsen 1 2 +Thomas_Birmingham 1 2 +Tiger_Woods 2 6 +Tiger_Woods 3 21 +Tiger_Woods 5 11 +Tiger_Woods 5 12 +Tiger_Woods 6 18 +Tiger_Woods 6 21 +Tiger_Woods 9 12 +Tiger_Woods 9 23 +Tiger_Woods 16 18 +Tim_Curry 1 2 +Tom_Brady 1 2 +Toshihiko_Fukui 1 2 +Toshihiko_Fukui 1 3 +Uma_Thurman 1 3 +Uma_Thurman 2 3 +Vaclav_Klaus 1 2 +Vanessa_Incontrada 1 2 +Vanessa_Incontrada 1 3 +Vanessa_Incontrada 1 4 +Vanessa_Incontrada 2 3 +Vanessa_Incontrada 2 4 +Vanessa_Incontrada 3 4 +Vin_Diesel 1 2 +Vladimir_Spidla 1 2 +Vladimir_Spidla 1 3 +Vladimir_Spidla 2 3 +Win_Aung 1 3 +Win_Aung 1 4 +Zhang_Ziyi 1 3 +Zhang_Ziyi 2 3 +Adrian_Annus 1 Jorge_Marquez-Ruarte 1 +Adrian_Annus 1 Patrick_Bourrat 1 +Adrian_Murrell 1 Jose_Cevallos 1 +Adrian_Murrell 1 Paul_Brandt 1 +Ahmed_Ibrahim_Bilal 1 Beatrice_Dalle 1 +Ahmed_Ibrahim_Bilal 1 Lee_Chang-dong 1 +Aileen_Riggin_Soule 1 Norio_Ohga 1 +Aitor_Gonzalez 2 Horace_Donovan_Reid 1 +Ajit_Agarkar 1 Jesse_James 1 +Akbar_Al_Baker 1 Andrei_Konchalovsky 1 +Akbar_Al_Baker 1 Bobby_Goldwater 1 +Alain_Cervantes 1 Bob_Huggins 3 +Alain_Cervantes 1 Pierre_Png 1 +Alanna_Ubach 1 Paul_Gascoigne 1 +Alec_Baldwin 4 Goh_Kun 2 +Alex_Holmes 1 Beatrice_Dalle 1 +Alfred_Sant 1 Randall_Tobias 1 +Alfredo_Moreno 1 Dyab_Abou_Jahjah 1 +Alfredo_Moreno 1 Suzanne_Torrance 1 +Alfredo_Pena 1 Emmanuel_Milingo 1 +Alvaro_Noboa 1 Phan_Van_Khai 1 +Alvaro_Noboa 2 Timothy_Rigas 1 +Amanda_Coetzer 1 Bridgette_Wilson-Sampras 2 +Amanda_Coetzer 1 Eddie_Jordan 1 +Amer_al-Saadi 4 Miguel_Cotto 1 +Amy_Brenneman 1 Russell_Coutts 2 +Amy_Brenneman 1 Terje_Roed-Larsen 1 +Andrew_Firestone 1 Harry_Schmidt 4 +Andrew_Firestone 1 Ian_Campbell 1 +Andrew_Firestone 1 Jose_Acasuso 1 +Andrew_Firestone 1 Omar_Sharif 4 +Andy_Graves 1 Gil_de_Ferran 1 +Angela_Alvarado_Rosa 1 Gerald_Calabrese 1 +Angela_Alvarado_Rosa 1 Jeffrey_Jones 1 +Anneli_Jaatteenmaki 2 Eric_Robert_Rudolph 2 +Anneli_Jaatteenmaki 2 Gerry_Adams 6 +Anneli_Jaatteenmaki 2 Jorge_Marquez-Ruarte 1 +Anthony_Corso 1 Johnny_Benson 1 +Anthony_Corso 1 Ray_Allen 3 +Anthony_Pico 1 Stephanie_Cohen_Aloro 1 +Ari_Fleischer 1 Razali_Ismail 1 +Ari_Fleischer 9 Jean-Marc_de_La_Sabliere 2 +Arianna_Huffington 4 Philippe_Gagnon 1 +Arnaud_Clement 1 Jeffrey_Jones 2 +Arnaud_Clement 1 King_Gyanendra 1 +Arnaud_Clement 2 Grady_Irvin_Jr 2 +Arnold_Scott 1 Randy_Dryer 1 +Art_Howe 2 Luciano_Bovicelli 1 +Art_Howe 3 Teresa_Worbis 1 +Artieas_Shanks 1 Derek_King 1 +Astrid_Betancourt 1 Frederick_Madden 1 +Barbara_De_Brun 1 Cosmo_Iacavazzi 1 +Barry_Collier 1 Narendra_Modi 1 +Ben_Affleck 3 Patricia_Hearst 1 +Ben_Affleck 4 Daniel_Scioli 1 +Ben_Affleck 7 Marcio_de_Souza 1 +Ben_Broussard 1 Rudi_Voeller 1 +Betsy_Smith 2 Steven_Kinlock 1 +Betty_Williams 1 Mary_Catherine_Correll 1 +Bianca_Jagger 1 Dario_Camuffo 1 +Bianca_Jagger 1 Fabricio_Oberto 1 +Bill_Callahan 1 Gina_Gershon 1 +Bill_Kollar 1 Gina_Gershon 1 +Bill_Kollar 1 Jose_Miguel_Aleman 1 +Bill_Kollar 1 Mae_Jemison 1 +Bill_Kollar 1 Mohammed_Ashraf_Hafiz 1 +Bill_Kollar 1 Patty_Sheehan 1 +Bill_Parsons 1 Donna_Barrera 1 +Bill_Stein 1 Vanessa_Incontrada 4 +Blythe_Hartley 1 Jackie_Dennis 1 +Blythe_Hartley 2 Gil_de_Ferran 2 +Bob_Curtis 1 Helen_Clark 1 +Bob_Huggins 3 Sylvester_Stallone 2 +Bobby_Goldwater 1 John_Moxley 1 +Bobby_Goldwater 1 Ulrich_Kueperkoch 1 +Bono 3 Ian_Huntley 1 +Brad_Garrett 1 Hoda_Asfor 1 +Brad_Garrett 4 Wang_Nan 1 +Brandon_Hammond 1 Thomas_Kelly 1 +Brandon_Robinson 1 Giovanny_Cordoba 1 +Brandon_Robinson 1 Michael_Linscott 1 +Brandon_Robinson 1 Shi_Guangsheng 1 +Brendan_Stai 1 Dan_Guerrero 1 +Brett_Boone 1 Jean-Marc_de_La_Sabliere 1 +Brett_Boone 1 Teresa_Worbis 1 +Brian_Billick 1 Ian_Huntley 1 +Brian_Billick 1 John_Wayne 1 +Brian_Billick 1 Stephanie_Moore 1 +Brian_Olson 1 Roberto_Tovar 1 +Bud_Selig 1 Franz_Fischler 1 +Bud_Selig 3 John_Duprey 1 +Carla_Moreno 1 Suzanne_Torrance 1 +Carla_Moreno 1 Tang_Jiaxuan 3 +Carlos_Beltran 1 Ulrich_Kueperkoch 1 +Carlos_Ghosn 1 Jim_Paxson 1 +Carlos_Ghosn 2 Ray_Allen 3 +Carlton_Dotson 1 Jim_Paxson 1 +Carlton_Dotson 1 Patrick_Bourrat 1 +Carlton_Dotson 1 Porter_Goss 1 +Carlton_Dotson 1 Vyacheslav_Fetisov 1 +Cass_Ballenger 1 Norio_Ohga 1 +Cecile_de_France 1 Dyab_Abou_Jahjah 1 +Cecile_de_France 1 Terrence_Kiel 1 +Charles_Bell 1 Tatjana_Gsell 1 +Charlotte_Rampling 1 Lana_Clarkson 2 +Chick_Hearn 1 Mohammed_Salmane 1 +Christopher_Reeve 4 Pauline_Landers 1 +Christopher_Reeve 4 Scott_OGrady 1 +Cindy_Margolis 1 Mark_Cuban 2 +Claire_Danes 2 Nadine_Vinzens 1 +Claudio_Lopez 1 Gabrielle_Rose 1 +Collis_Temple_III 1 Eva_Amurri 1 +Collis_Temple_III 1 Rob_Niedermayer 1 +Collis_Temple_III 1 Santiago_Botero 1 +Collis_Temple_III 1 Simon_Chalk 1 +Conan_OBrien 1 Liliana_Cavani 1 +Corinna_Harfouch 1 Ivo_Dubs 1 +Corinna_Harfouch 1 Tim_Curry 1 +Cristina_Kirchner 1 Stefanie_De_Roux 1 +Dan_Boyle 1 Paul_Clark 1 +Dan_Boyle 1 Thabo_Mbeki 3 +Daniel_Chin 1 Ian_Huntley 1 +Daniel_Chin 1 Jim_Letten 1 +Daniel_Chin 1 Julia_Glass 1 +Daniel_Scioli 1 Lena_Katina 1 +Daniel_Scioli 1 Lindsay_Lohan 1 +Dario_Camuffo 1 Eli_Stutsman 1 +David_Brinkley 1 Jeff_Bridges 1 +David_Brinkley 1 Stephen_Arigbabu 1 +David_Montoya 1 Mary_Elizabeth_Mastrantonio 1 +David_Siegel 1 Francis_Mer 1 +Dennis_Johnson 1 Satnarine_Sharma 1 +Denys_Arcand 1 Nadine_Vinzens 2 +Derek_Jeter 1 Tian_Zhuang_Zhuang 1 +Derek_Jeter 3 Jolanta_Kwasniewski 2 +Derek_King 1 Yasein_Taher 1 +Derrick_Battie 1 Ian_Huntley 1 +Diego_Armando_Maradona 1 Robert_Gordon_Card 1 +Don_King 1 Thomas_Kelly 1 +Don_King 1 Zurab_Tsereteli 1 +Donna_Barrera 1 Francis_Mer 2 +Donna_Shalala 2 Sereyvuth_Kem 1 +Duncan_Fletcher 1 John_Cusack 2 +Duncan_Fletcher 1 Mike_Alden 1 +Dustin_Brown 1 Jose_Cevallos 1 +Dyab_Abou_Jahjah 1 John_Cornyn 1 +Ed_Mekertichian 1 Paul_Clark 1 +Edmund_Hillary 2 Miguel_Cotto 1 +Elena_Likhovtseva 1 Marty_Mornhinweg 2 +Eli_Stutsman 1 Kalpana_Chawla 4 +Eli_Stutsman 1 Marcio_de_Souza 1 +Eli_Stutsman 1 Michael_Phelps 1 +Elizabeth_Regan 1 Eugene_Teslovic 1 +Emmanuel_Milingo 1 Enrica_Fico 1 +Enola_Rice 1 Lana_Clarkson 2 +Erin_Runnion 3 Rob_Lowe 2 +Eugene_Teslovic 1 Goh_Kun 1 +Eugene_Teslovic 1 Jean-Marc_de_La_Sabliere 2 +Eva_Amurri 1 Tanya_Holyk 1 +Fabricio_Oberto 1 Rudi_Voeller 1 +Felix_Sanchez 1 Ian_Moran 1 +Fernando_Henrique_Cardoso 1 Michael_Phelps 1 +Fernando_Henrique_Cardoso 8 Sherry_Fisher 1 +Francis_Mer 2 Patrick_Bourrat 1 +Franz_Fischler 2 Robert_Lange 1 +Frederick_Madden 1 Gary_Condit 1 +Gary_Bauer 1 Greg_Hodge 1 +Gary_Condit 1 Nathan_Lane 2 +Gary_Locke 1 Marcio_de_Souza 1 +Gen_Meredith 1 Matthew_Vaughan 1 +Gerry_Adams 8 Jamir_Miller 1 +Gilles_Panizzi 1 Jim_Piper 1 +Gina_Gershon 1 Patricia_Garone 1 +Gina_Gershon 1 Zhang_Ziyi 2 +Giovanny_Cordoba 1 Jackie_Sherrill 1 +Giovanny_Cordoba 1 Keith_Olbermann 1 +Grady_Irvin_Jr 2 Oracene_Williams 1 +Graeme_Smith 1 John_Salazar 1 +Graham_Bentley 1 Jason_Clermont 1 +Guillaume_Depardieu 1 Manuel_Llorente 1 +Gunter_Pleuger 1 Nicole_Kidman 2 +Gunter_Pleuger 2 Robby_Ginepri 1 +Gunter_Pleuger 3 Vaclav_Klaus 1 +Hanns_Schumacher 1 Martin_Gecht 1 +Henri_Proglio 1 Ibrahim_Haddad 1 +Henri_Proglio 1 Jeffery_Hendren 1 +Hermann_Maier 1 Phan_Van_Khai 1 +Hermann_Maier 2 Stephen_Arigbabu 1 +Hideki_Sato 1 Peter_Mugyeni 1 +Hideki_Sato 1 Troy_Aikman 1 +Hoda_Asfor 1 Juergen_Peters 1 +Hoda_Asfor 1 Lindsay_Lohan 1 +Horace_Donovan_Reid 1 Rob_Lowe 1 +Hunter_Kemper 1 Marsha_Thomason 1 +Ian_Campbell 1 Mike_Alden 1 +Ian_Huntley 1 Jalen_Rose 1 +Ian_Huntley 1 Sebastian_Saja 2 +Ian_Moran 1 Stephanie_Cohen_Aloro 1 +Ian_Thorpe 5 Mickey_Gilley 1 +Ignacio_Antonio_Velasco 1 Rich_Brooks 1 +Iran_Brown 1 Margaret_Caruso 1 +Isabelle_Huppert 1 Marcio_de_Souza 1 +Isabelle_Huppert 2 Phan_Van_Khai 3 +Ivan_Shvedoff 1 Josh_Childress 1 +Ivan_Shvedoff 1 Miguel_Estrada 2 +Ivan_Shvedoff 1 Vojislav_Seselj 1 +Izzat_Ibrahim 1 Jerry_Rice 1 +Jacques_Villeneuve 1 Wim_Duisenberg 1 +Jalen_Rose 1 Suzanne_Torrance 1 +James_Butts 1 Nicole_Kidman 2 +James_Butts 2 Jerry_Colangelo 1 +James_Butts 2 Sherry_Fisher 1 +James_May 1 Trevor_McDonald 1 +James_May 1 Vaclav_Klaus 2 +Jamie_Cooke 1 Kalpana_Chawla 5 +Jamie_Cooke 1 Mickey_Rooney 1 +Jamir_Miller 1 Laura_Hernandez 1 +Jana_Pittman 1 Liv_Tyler 1 +Jason_Clermont 1 Jose_Carlo_Fernandez 1 +Jason_Clermont 1 Patty_Duke 1 +Jeff_Bridges 1 Landon_Donovan 1 +Jeff_Bridges 1 Patty_Duke 1 +Jeffery_Hendren 1 Jeremy_Wotherspoon 1 +Jennifer_McCoy 1 Manuel_Llorente 1 +Jennifer_McCoy 1 William_Cocksedge 1 +Jerry_Colangelo 1 Vin_Diesel 2 +Jerry_Rice 1 Joe_Gatti 2 +Jessica_Brungo 1 Landon_Donovan 1 +Jim_OBrien 1 Nadine_Vinzens 1 +Jim_Paxson 1 Robert_Gordon_Card 1 +Joe_Gatti 2 Mike_Samp 1 +Joel_Todd 1 John_Fox 1 +Joel_Todd 1 Momir_Nikolic 1 +John_Cornyn 1 Jose_Mourinho 2 +John_Cornyn 1 Robert_Weitzel 1 +John_Cornyn 1 Toshihiko_Fukui 2 +John_Dallager 1 Jose_Canseco_Sr 1 +John_Spencer 2 Stephanie_Moore 1 +John_Walsh 1 Kalpana_Chawla 1 +John_Wright 1 Sandra_Milo 1 +John_Wright 1 Trevor_McDonald 1 +Johnny_Benson 1 Lana_Clarkson 2 +Johnny_Benson 1 Teresa_Worbis 1 +Jolanta_Kwasniewski 2 Martin_Howard 1 +Jose_Canseco_Sr 1 Julia_Glass 1 +Jose_Mourinho 1 Joseph_LePore 1 +Jose_Mourinho 2 LeRoy_Millette_Jr 1 +Jose_Vicente_Rangel 1 Leuris_Pupo 1 +Jose_Vicente_Rangel 1 Scott_OGrady 1 +Joy_Lee_Sadler 1 Laurie_Chan 1 +Joy_Lee_Sadler 1 Norio_Ohga 1 +Joy_Lee_Sadler 1 Stephanie_Moore 1 +Karen_Pereiras 1 Michael_Phelps 2 +Keith_Brown 1 Nicole_Kidman 17 +Keith_Brown 1 William_Harrison 1 +Keith_Van_Horn 1 Sherry_Fisher 1 +Kelli_White 1 Rudi_Voeller 1 +King_Gyanendra 1 Otto_Reich 1 +Lana_Clarkson 1 Mike_Samp 1 +Landon_Donovan 1 Robby_Ginepri 1 +Larry_Tanenbaum 1 Mike_Samp 1 +Laura_Ziskin 1 Reyyan_Uzuner 1 +Laura_Ziskin 1 Robert_Gordon_Card 1 +LeRoy_Millette_Jr 1 Leuris_Pupo 1 +Lee_Chang-dong 1 Phil_Bredesen 1 +Liliana_Cavani 1 Richard_Pennington 1 +Lindsay_Lohan 1 Mireya_Elisa_Moscoso_Rodriguez 1 +Lloyd_Ward 2 Tina_Andrews 1 +Mark_Heller 1 Nicolas_Kiefer 1 +Mark_Heller 1 Parris_Glendening 1 +Mark_Heller 1 Peter_Hillary 2 +Martin_Gecht 1 Peter_Hillary 1 +Mary_Elizabeth_Mastrantonio 1 Vaclav_Klaus 1 +Michael_Arif 1 Sean_Patrick_OMalley 3 +Michael_Linscott 1 Tom_Brady 2 +Mick_Jagger 3 Oracene_Williams 1 +Mo_Elleithee 1 Tanya_Holyk 1 +Mohammed_Ashraf_Hafiz 1 Rod_Bryden 1 +Mukhtar_Alytnbayev 1 Oracene_Williams 1 +Patricia_Garone 1 Sean_Patrick_OMalley 2 +Patricia_Hearst 1 Scott_Dickson 1 +Patty_Duke 1 Simon_Chalk 1 +Paul_Bettany 1 Ulrich_Kueperkoch 1 +Paul_Gascoigne 3 Tian_Liang 1 +Paul_Greengrass 1 Tim_Pawlenty 1 +Perry_Compton 1 William_Shatner 1 +Peter_Camejo 1 Ruth_Christofferson 1 +Porter_Goss 1 Tara_Kirk 1 +Rand_Miller 1 Robert_Nillson 1 +Rich_Brooks 1 Sharess_Harrell 1 +Richard_Pennington 1 Robby_Ginepri 2 +Robert_Gordon_Card 1 William_Shatner 1 +Robert_Kipkoech_Cheruiyot 1 Sally_Kirkland 3 +Robert_Kipkoech_Cheruiyot 1 Steve_Park 1 +Russell_Coutts 1 Tian_Liang 1 +Sandra_Milo 1 Satnarine_Sharma 1 +Steve_Nesbitt 1 Win_Aung 4 +Sylvester_Stallone 7 TJ_Ford 1 +Sylvie_Guillem 1 Vadim_Devyatovskiy 1 +Tara_Kirk 1 Win_Aung 1 +Aaron_Peirsol 1 4 +Aaron_Peirsol 3 4 +Adrian_Nastase 1 2 +Ahmed_Chalabi 1 3 +Ahmed_Chalabi 1 5 +Albrecht_Mentz 1 2 +Alejandro_Toledo 20 36 +Alejandro_Toledo 21 24 +Alejandro_Toledo 21 30 +Alejandro_Toledo 23 27 +Alejandro_Toledo 26 29 +Alexander_Losyukov 1 3 +Alexander_Losyukov 2 3 +Alexander_Losyukov 2 4 +Alimzhan_Tokhtakhounov 1 2 +Amelie_Mauresmo 7 14 +Amelie_Mauresmo 11 17 +Amelie_Mauresmo 14 17 +Angelo_Reyes 1 2 +Angelo_Reyes 1 3 +Begum_Khaleda_Zia 1 2 +Ben_Curtis 1 2 +Bijan_Namdar_Zangeneh 1 2 +Bill_Paxton 1 2 +Bill_Paxton 1 3 +Bill_Paxton 2 4 +Bill_Paxton 3 4 +Billy_Crystal 1 2 +Billy_Crystal 1 3 +Billy_Crystal 1 5 +Billy_Crystal 3 5 +Billy_Graham 1 2 +Bob_Colvin 1 2 +Brian_Cowen 1 2 +Butch_Davis 1 2 +Byron_Scott 1 2 +Carol_Burnett 1 2 +Charles_Mathews 1 2 +Christine_Ebersole 1 2 +Claudia_Schiffer 1 2 +Condoleezza_Rice 1 4 +Condoleezza_Rice 2 10 +Condoleezza_Rice 4 5 +Condoleezza_Rice 8 9 +Condoleezza_Rice 9 10 +Costas_Simitis 1 4 +Costas_Simitis 1 5 +Costas_Simitis 3 5 +Costas_Simitis 4 6 +Cristina_Saralegui 1 2 +Cruz_Bustamante 1 4 +Cruz_Bustamante 3 5 +David_Heymann 3 5 +Diana_Krall 1 6 +Diana_Krall 3 4 +Diana_Munz 1 3 +Diana_Munz 2 3 +Dominique_de_Villepin 2 13 +Dominique_de_Villepin 3 8 +Dominique_de_Villepin 3 10 +Dominique_de_Villepin 7 11 +Dominique_de_Villepin 10 14 +Dominique_de_Villepin 11 12 +Don_Siegelman 1 3 +Don_Siegelman 2 4 +Donald_Pettit 1 2 +Donald_Pettit 1 3 +Donald_Pettit 2 3 +Doug_Collins 1 2 +Edward_James_Olmos 1 2 +Elin_Nordegren 1 2 +Elizabeth_Taylor 1 2 +Ellen_DeGeneres 1 2 +Elton_John 1 4 +Elton_John 2 3 +Elton_John 3 6 +Elton_John 5 6 +Elton_John 5 7 +Emma_Watson 2 3 +Emma_Watson 3 4 +Emma_Watson 3 5 +Fabrice_Santoro 1 3 +Fabrice_Santoro 2 3 +Flavia_Delaroli 1 2 +George_Lopez 1 4 +George_Lopez 2 4 +George_Lopez 2 5 +George_Lopez 4 5 +Gerard_Depardieu 1 2 +Gerhard_Schroeder 5 43 +Gerhard_Schroeder 17 25 +Gerhard_Schroeder 27 32 +Gerhard_Schroeder 45 67 +Gerhard_Schroeder 66 100 +Gilberto_Rodriguez_Orejuela 1 3 +Gilberto_Rodriguez_Orejuela 2 3 +Gilberto_Rodriguez_Orejuela 3 4 +Giuseppe_Gibilisco 1 3 +Giuseppe_Gibilisco 1 4 +Giuseppe_Gibilisco 3 4 +Guillermo_Canas 1 2 +Guillermo_Canas 1 3 +Guillermo_Canas 2 4 +Guillermo_Canas 3 4 +Hans_Blix 7 8 +Hans_Blix 7 18 +Hans_Blix 7 38 +Hans_Blix 8 24 +Hans_Blix 9 12 +Hans_Blix 26 36 +Hans_Blix 27 37 +Hans_Blix 31 37 +Heath_Ledger 1 2 +Heath_Ledger 1 4 +Heath_Ledger 2 3 +Heath_Ledger 2 4 +Herta_Daeubler-Gmelin 1 2 +Hilary_Duff 1 3 +Hilary_Duff 2 3 +Hosni_Mubarak 6 7 +Hun_Sen 1 4 +Hun_Sen 2 4 +Iain_Duncan_Smith 1 3 +Iain_Duncan_Smith 2 3 +Iain_Duncan_Smith 2 4 +Iain_Duncan_Smith 3 4 +Inocencio_Arias 1 2 +JK_Rowling 1 2 +JK_Rowling 2 3 +JK_Rowling 2 4 +JK_Rowling 3 5 +Jackie_Chan 1 3 +Jackie_Chan 5 8 +Jane_Kaczmarek 1 2 +Janet_Napolitano 1 2 +Janet_Napolitano 1 4 +Jay_Leno 1 3 +Jay_Leno 2 3 +Jean-David_Levitte 1 2 +Jean-David_Levitte 1 4 +Jean-David_Levitte 2 5 +Jean-David_Levitte 3 4 +Jean-David_Levitte 4 9 +Jeff_Van_Gundy 1 2 +Jeff_Van_Gundy 1 3 +Jeff_Van_Gundy 2 3 +Jim_Furyk 1 6 +Jim_Furyk 3 6 +John_Garamendi 1 2 +John_Rigas 1 2 +John_Warner 1 2 +John_Warner 1 3 +John_Warner 2 4 +John_Warner 3 4 +Julia_Tymoshenko 1 2 +Julia_Tymoshenko 1 3 +Julia_Tymoshenko 2 3 +Keith_Bogans 1 3 +Keith_Bogans 2 3 +Ken_Macha 1 3 +Ken_Macha 2 3 +Kenneth_Bowersox 1 2 +Kenneth_Bowersox 1 3 +Kenneth_Bowersox 2 3 +Kristanna_Loken 1 2 +Kristanna_Loken 3 4 +Kristanna_Loken 4 5 +LK_Advani 1 2 +LK_Advani 1 3 +Lauren_Killian 1 2 +Lennox_Lewis 1 2 +Lennox_Lewis 1 3 +Lili_Taylor 1 2 +Lily_Tomlin 1 2 +Lon_Kruger 1 2 +Lynn_Abraham 1 2 +Mahmoud_Abbas 7 11 +Mahmoud_Abbas 15 20 +Mahmoud_Abbas 22 27 +Mark_Richt 1 2 +Mark_Richt 1 3 +Mary-Kate_Olsen 1 2 +Mary-Kate_Olsen 2 3 +Matt_Dillon 1 2 +Max_Mayfield 1 2 +Megan_Mullally 1 2 +Megan_Mullally 1 3 +Michael_Caine 1 2 +Michael_Caine 1 3 +Michael_Caine 3 4 +Michael_J_Sheehan 1 2 +Michael_Keaton 1 2 +Michael_Powell 1 2 +Michael_Powell 1 3 +Michael_Powell 1 5 +Michael_Powell 2 3 +Michael_Powell 2 4 +Michael_Powell 2 5 +Michael_Powell 4 5 +Miguel_Contreras 1 2 +Mike_Weir 1 6 +Mike_Weir 3 11 +Mikhail_Kasyanov 1 2 +Mikhail_Kasyanov 1 3 +Mikhail_Kasyanov 3 4 +Mohammed_Baqir_al-Hakim 1 3 +Monica_Lewinsky 1 2 +Monica_Lewinsky 1 3 +Monica_Lewinsky 2 3 +Nan_Wang 1 3 +Nan_Wang 2 3 +Nan_Wang 3 4 +Nancy_Demme 1 2 +Nancy_Reagan 1 2 +Naoto_Kan 1 3 +Naoto_Kan 2 3 +Natasha_McElhone 1 2 +Natasha_McElhone 1 3 +Neri_Marcore 1 2 +Nicholas_Tse 1 2 +Nicolas_Escude 1 2 +Noelle_Bush 1 3 +Noelle_Bush 1 4 +Noelle_Bush 2 4 +Orrin_Hatch 1 2 +Padraig_Harrington 1 3 +Padraig_Harrington 1 4 +Padraig_Harrington 2 4 +Pedro_Almodovar 1 2 +Pedro_Almodovar 3 5 +Pedro_Almodovar 5 6 +Penelope_Ann_Miller 1 2 +Peter_Arnett 1 3 +Peter_Arnett 2 3 +Petria_Thomas 1 2 +Petro_Symonenko 1 2 +Prince_Charles 1 2 +Prince_Charles 3 4 +Prince_Charles 3 5 +Raoul_Ruiz 1 2 +Raoul_Ruiz 2 3 +Raoul_Ruiz 2 4 +Raoul_Ruiz 3 4 +Ricky_Ponting 1 2 +Robert_Stack 1 2 +Robin_Cook 1 2 +Roger_Federer 4 5 +Roger_Federer 4 7 +Roger_Federer 6 7 +Ronaldo_Luis_Nazario_de_Lima 1 4 +Ronaldo_Luis_Nazario_de_Lima 2 3 +Ronaldo_Luis_Nazario_de_Lima 2 4 +Ronaldo_Luis_Nazario_de_Lima 3 4 +Roy_Williams 1 3 +Roy_Williams 2 4 +Sally_Ride 1 2 +Scott_McNealy 1 2 +Scott_Peterson 1 2 +Scott_Peterson 1 5 +Scott_Peterson 2 5 +Scott_Peterson 3 4 +Scott_Ritter 1 2 +Sean_OKeefe 3 4 +Sean_Penn 1 2 +Sean_Penn 2 3 +Sebastien_Grosjean 1 2 +Sebastien_Grosjean 1 4 +Sharon_Stone 1 2 +Sharon_Stone 1 3 +Sharon_Stone 1 4 +Sharon_Stone 1 5 +Sharon_Stone 2 4 +Sonia_Gandhi 1 3 +Sonia_Gandhi 1 4 +Steve_Backley 1 2 +Steven_Hatfill 1 2 +Tassos_Papadopoulos 1 3 +Taufik_Hidayat 1 2 +Taufik_Hidayat 1 3 +Taufik_Hidayat 2 3 +Thomas_Malchow 1 2 +Tom_Cruise 3 5 +Tom_Cruise 4 10 +Tom_Cruise 6 10 +Toni_Braxton 1 3 +Wayne_Gretzky 1 2 +Wayne_Gretzky 1 4 +William_Bulger 1 3 +William_Bulger 2 4 +Winona_Ryder 12 13 +Winona_Ryder 12 18 +Winona_Ryder 12 19 +Xanana_Gusmao 1 4 +Xanana_Gusmao 2 3 +Xanana_Gusmao 3 5 +Xanana_Gusmao 4 5 +Yann_Martel 1 2 +Yashwant_Sinha 1 3 +Yashwant_Sinha 1 5 +Yashwant_Sinha 5 7 +Aaron_Guiel 1 Pascal_Rheaume 1 +Aaron_Guiel 1 Steve_Zahn 1 +Adrian_Nastase 2 Princess_Victoria 1 +Ahmed_Chalabi 1 Rosario_Dawson 1 +Ahmed_Ghazi 1 Julia_Tymoshenko 3 +Ahmet_Demir 1 Wally_Szczerbiak 1 +Ain_Seppik 1 Gerhard_Schroeder 55 +Ain_Seppik 1 Misty_Dawn_Clymer 1 +Ain_Seppik 1 Raoul_Ruiz 1 +Ain_Seppik 1 Susan_Whelan 1 +Aishwarya_Rai 1 Chloe_Sevigny 1 +Aishwarya_Rai 1 Dinah_Turner 1 +Aishwarya_Rai 1 Rosario_Dawson 1 +Alan_Zemaitis 1 Lauren_Killian 1 +Alex_Cejka 1 Taufik_Hidayat 2 +Alexandre_Daigle 1 Brent_Coles 1 +Ali_Bin_Hussein 1 Ben_Cahoon 1 +Ali_Bin_Hussein 1 Oswald_Gruebel 1 +Ali_Bin_Hussein 1 Raf_Vallone 1 +Alimzhan_Tokhtakhounov 1 Jeane_Kirkpatrick 1 +Alimzhan_Tokhtakhounov 2 Butch_Davis 1 +Alisha_Richman 1 Camille_Lewis 1 +Alisha_Richman 1 Katerina_Smrzova 1 +Alisha_Richman 1 Miguel_Rosseto 1 +Amelie_Mauresmo 3 Laura_Flessel 1 +Amelie_Mauresmo 18 Bob_Colvin 1 +Ana_Claudia_Talancon 1 Piers_Sellers 1 +Andrea_Kiser 1 Ellen_Barkin 1 +Andrea_Kiser 1 Imre_Kertasz 1 +Andrea_Kiser 1 Matt_Dillon 1 +Andrea_Kiser 1 Yann_Martel 2 +Andres_DAlessandro 1 Leticia_Van_de_Putte 1 +Andres_DAlessandro 1 Ronaldo_Luis_Nazario_de_Lima 1 +Anette_Hosoi 1 Rolf_Zimmermann 1 +Anette_Hosoi 1 Sheila_Taormina 1 +Anette_Hosoi 1 Tora_Takagi 1 +Angelo_Reyes 1 Chip_Knight 1 +Angelo_Reyes 1 Oliver_Neuville 1 +Anne_Heche 1 Sasha_Alexander 1 +Anne_Heche 1 Tanya_Lindenmuth 1 +Antje_Buschschulte 1 Jerry_Angelo 1 +Anton_Balasingham 1 Matt_Siebrandt 1 +Anton_Balasingham 1 Will_Ferrell 1 +Armando_Avila_Panchame 1 Monica_Gabrielle 1 +Armando_Avila_Panchame 1 Retief_Goosen 1 +BJ_Habibie 1 Craig_MacTavish 1 +Baz_Luhrmann 1 Danny_Avalon 1 +Begum_Khaleda_Zia 1 Eric_Idle 1 +Begum_Khaleda_Zia 1 Michael_Andretti 1 +Begum_Khaleda_Zia 1 Picabo_Street 1 +Ben_Cahoon 1 Nate_Blackwell 1 +Ben_Cahoon 1 Roy_Williams 3 +Ben_Curtis 3 Jim_Calhoun 1 +Benicio_Del_Toro 1 Elizabeth_Taylor 1 +Benicio_Del_Toro 1 Ismail_Khan 1 +Benicio_Del_Toro 1 Tonya_Payne 1 +Bijan_Namdar_Zangeneh 1 Petro_Symonenko 1 +Billy_Graham 1 Karin_Pilsaeter 1 +Billy_Graham 1 Piers_Sellers 1 +Billy_Graham 2 Howard_Wilkinson 1 +Bob_Colvin 2 Janet_Napolitano 1 +Boris_Henry 1 John_Reid 2 +Brandon_Boyd 1 James_Brosnahan 1 +Brandon_Boyd 1 Neil_Moritz 1 +Brent_Coles 1 Zoe_Ball 1 +Brett_Hull 1 Gregorio_Rosal 1 +Brett_Hull 1 Kirsten_Clark 1 +Brett_Hull 1 Ralph_Nader 1 +Brett_Perry 1 Neri_Marcore 1 +Bruce_Willis 1 Carlos_Juarez 1 +Bruce_Willis 1 Jim_Carrey 1 +Bruce_Willis 1 Petria_Thomas 2 +Bryant_Young 1 Jim_Bollman 1 +Bryant_Young 1 Maurice_Cheeks 1 +Buddy_Ryan 1 James_Brosnahan 1 +Buddy_Ryan 1 Nancy_Reagan 2 +Butch_Davis 2 Herta_Daeubler-Gmelin 2 +Byron_Scott 1 Jane_Clayson 1 +Camille_Lewis 1 Lincoln_Chafee 1 +Camille_Lewis 1 Nathirah_Hussein 1 +Carol_Burnett 1 Raf_Vallone 1 +Carol_Niedermayer 1 Kristin_Scott 1 +Catherine_Ndereba 1 Janet_Napolitano 4 +Chandrika_Kumaratunga 1 Kevin_Satterfield 1 +Charles_Mathews 1 Will_Ferrell 1 +Charles_Mathews 2 Kristin_Scott 1 +Charlie_Garner 1 Luo_Linquan 1 +Charlie_Williams 1 Julia_Tymoshenko 1 +Cheryl_Ford 1 Chris_Cookson 1 +Chip_Knight 1 Ronald_Post 1 +Chloe_Sevigny 1 Emyr_Jones_Parry 1 +Chloe_Sevigny 1 Petria_Thomas 3 +Chris_Cornell 1 Desmon_Farmer 1 +Christine_Ebersole 2 Paul_Walker 1 +Christoph_Daum 1 Scott_Hubbard 1 +Christopher_Conyers 1 Michael_Milton 1 +Christopher_Conyers 1 Raoul_Ruiz 1 +Chuck_Bednarik 1 Mario_Jardel 1 +Cindy_Zagorski 1 Rosie_Perez 1 +Claudia_Schiffer 1 Lynn_Abraham 2 +Claudia_Schiffer 3 Lin_Yung_Hsi 1 +Colin_Cowie 1 Mstislav_Rostropovich 1 +Condoleezza_Rice 2 Mehmet_Okur 1 +Condoleezza_Rice 7 Ryan_Nyquist 1 +Costas_Simitis 2 Lorraine_Fenton 1 +Craig_MacTavish 1 Vince_Dooley 1 +Cristina_Saralegui 1 Dave_Robertson 1 +Cruz_Bustamante 2 Dean_Barkley 3 +Cyndi_Thompson 1 Nicolas_Escude 2 +Cyndi_Thompson 2 Owen_Nolan 1 +Dalia_Rabin-Pelosoff 1 Sean_Combs 1 +Daniel_Darnell 1 Malcolm_Jamal_Warner 1 +Danny_Avalon 1 Sean_Penn 2 +David_Blaine 1 Nicholas_Tse 2 +David_Blaine 1 Ramon_Delgado 1 +David_Heymann 3 Norm_Macdonald 1 +David_Surrett 1 Michael_Taylor 1 +David_Tornberg 1 Ricky_Ponting 1 +Deece_Eckstein 1 Diana_Munz 1 +Deece_Eckstein 1 Tommy_Lewis 1 +Denis_Coderre 1 Don_Siegelman 1 +Denis_Coderre 1 John_Eder 1 +Desmon_Farmer 1 Lin_Yung_Hsi 1 +Diana_Krall 4 Ramon_Delgado 1 +Dick_Armey 1 Mehmet_Okur 1 +Dinah_Turner 1 Mekhi_Phifer 1 +Dinah_Turner 1 Tammy_Helm 1 +Don_Boudria 1 Roy_Chaderton 1 +Don_Boudria 1 Sananda_Maitreya 1 +Don_Matthews 1 Kirsten_Clark 1 +Don_Matthews 1 Pa_Kou_Hang 1 +Don_Matthews 1 Peter_Arnett 2 +Don_Siegelman 2 Elizabeth_Taylor 2 +Donald_Pettit 1 Edward_James_Olmos 2 +Doug_Collins 1 Kristanna_Loken 5 +Dyana_Calub 1 Ronald_Post 1 +Earl_Counter 1 Joe_Friedberg 1 +Edward_Albee 1 Nicola_Bono 1 +Edward_Belvin 1 Jacqueline_Gold 1 +Edward_James_Olmos 2 Mariano_Zabaleta 1 +Elin_Nordegren 1 George_Brumley_III 1 +Elin_Nordegren 1 Samantha_Daniels 1 +Elin_Nordegren 2 Leszek_Miller 3 +Elisabeth_Welch 1 Piers_Sellers 1 +Elizabeth_Taylor 1 Olympia_Dukakis 1 +Ellen_DeGeneres 2 Lorraine_Fenton 1 +Elva_Hsiao 1 Samantha_Daniels 1 +Emma_Watson 2 Henry_Hyde 1 +Emyr_Jones_Parry 1 Sally_Clark 1 +Eric_Ryan_Donnelly 1 Iain_Duncan_Smith 1 +Ernesto_Zedillo 1 Lenny_Kravitz 1 +Ernesto_Zedillo 1 Megan_Mullally 2 +Erwin_Mapasseng 1 Hilary_Duff 3 +Evelyn_Lauder 1 Nathan_Powell 1 +Fernando_Hierro 1 Prince_Charles 2 +Flavia_Delaroli 2 John_Warner 4 +Flavia_Delaroli 2 Vanessa_Laine 1 +Fran_Drescher 1 Warren_Truss 1 +Fran_Drescher 2 Jay_Leno 3 +Frank_Murkowski 1 James_Brosnahan 1 +Frank_Stallone 1 Tommy_Lewis 1 +Frank_Stallone 2 Toni_Braxton 2 +Gary_Coleman 1 Koichi_Tanaka 1 +Gary_Coleman 1 Uri_Lopolianski 1 +Gary_Sayler 1 Ralph_Sampson 1 +George_Brumley_III 1 Will_Ferrell 1 +George_Lopez 5 Iain_Duncan_Smith 1 +Georgina_Bardach 1 Henry_Hilow 1 +Gilberto_Rodriguez_Orejuela 3 Lennox_Lewis 1 +Gilberto_Rodriguez_Orejuela 3 Nancy_Demme 1 +Gordon_Lightfoot 1 Lima_Azimi 1 +Gordon_Lightfoot 1 Robert_Nardelli 1 +Gracia_Burnham 1 Jim_Calhoun 1 +Gregorio_Rosal 1 Richard_Greenberg 1 +Guenter_Verheugen 1 Kaspar_Villiger 1 +Guillermo_Canas 1 Tim_Duncan 2 +Guillermo_Canas 2 Michalis_Chrisohoides 1 +Hal_McCoy 1 Rudolph_Holton 1 +Hank_Aaron 1 Steve_Valentine 1 +Hans_Eichel 1 Jimmy_Iovine 1 +Hashan_Tillakaratne 1 John_Warner 3 +Heath_Ledger 2 Melissa_Stark 1 +Henry_Hilow 1 Monica_Gabrielle 1 +Herta_Daeubler-Gmelin 2 Koichi_Tanaka 1 +Hilary_Duff 1 Luther_Htu 1 +Hun_Sen 2 John_Rigas 2 +Idi_Amin 1 Stephen_Silas 1 +Isabel_Orellana 1 Spike_Helmick 1 +Ismael_Miranda 1 Janela_Jara 1 +Ismael_Miranda 1 Peter_Greenspun 1 +Ismail_Khan 1 Malcolm_Jamal_Warner 1 +JK_Rowling 5 Sasha_Alexander 1 +Jackie_Chan 3 Jennifer_Gratz 1 +Jacqueline_Edwards 1 Meg_Wakeman 1 +Jacqueline_Edwards 1 Noor_Mohammed 1 +James_Baker 1 Mike_Tice 1 +James_Harris 1 Keith_Bogans 2 +James_Layug 1 Ralph_Sampson 1 +Jane_Clayson 1 Rene_Antonio_Leon_Rodriguez 1 +Jane_Kaczmarek 2 Morris_Watts 1 +Jane_Russell 1 Martin_ONeill 1 +Janela_Jara 1 Kirsten_Clark 1 +Janela_Jara 1 Pa_Kou_Hang 1 +Janela_Jara 1 Ricky_Cottrill 1 +Janet_Napolitano 2 Laurie_Laychak 1 +Javier_Camara 1 Nate_Blackwell 1 +Jeane_Kirkpatrick 1 Jeff_Van_Gundy 3 +Jeane_Kirkpatrick 1 Richard_Greenberg 1 +Jeff_Van_Gundy 1 Richard_Naughton 1 +Jeffrey_Katzenberg 1 Susan_Whelan 1 +Jennette_Bradley 1 Tammy_Helm 1 +Jim_Furyk 6 Linn_Thornton 1 +John_Eastman 1 Ricky_Ponting 1 +John_F_Kennedy_Jr 2 Nathan_Powell 1 +John_Garamendi 1 Petro_Symonenko 2 +John_Madden 1 Sara_Silverman 1 +Jose_Luis_Rodriguez_Zapatero 1 Rene_Antonio_Leon_Rodriguez 1 +Jose_Luis_Rodriguez_Zapatero 1 Steve_Lenard 1 +Jose_Santos 1 Michael_Boyce 1 +Joseph_Kabila 1 Rose_Linkins 1 +Juan_Francisco_Palencia 1 Kristanna_Loken 2 +Juan_Francisco_Palencia 1 Warren_Truss 1 +Juan_Roman_Riquelme 1 Ralph_Sampson 1 +Julia_Tymoshenko 2 Luis_Sanchez 1 +Karin_Pilsaeter 1 Mike_Flanagan 1 +Karin_Pilsaeter 1 Petria_Thomas 3 +Keith_Bogans 2 Martin_ONeill 1 +Kelly_Leigh 1 Spike_Helmick 1 +Ken_Macha 3 Tommy_Lewis 1 +Kenneth_Carlsen 1 Robin_Cook 1 +Kevin_Borseth 1 Michael_Caine 4 +Koichi_Tanaka 1 Ricardo_Lopez_Murphy 1 +Kristin_Scott 1 Max_Mayfield 2 +Laura_Flessel 1 Owen_Nolan 1 +Lauren_Killian 2 Tommy_Lewis 1 +Laurie_Laychak 1 Pa_Kou_Hang 1 +Lawrence_Di_Rita 1 Mekhi_Phifer 1 +Lennox_Lewis 3 Monte_Kiffin 1 +Leon_Lai 1 Monica_Lewinsky 1 +Leon_Lai 1 Sun_Myung_Moon 1 +Lily_Tomlin 1 Marcos_Daniel_Jimenez 1 +Lily_Tomlin 2 Sim_Yong 1 +Linn_Thornton 1 Sherri_Coale 1 +Linn_Thornton 1 Tom_Cruise 2 +Lloyd_Mudiwa 1 Sebastien_Grosjean 1 +Lois_Smart 1 Tavis_Smiley 1 +Luis_Sanchez 1 Petro_Symonenko 1 +Luo_Linquan 1 Martin_ONeill 1 +Luther_Htu 1 Steve_Karsay 1 +Lynn_Abraham 1 Michael_Keaton 2 +Lynn_Abraham 2 Mariano_Zabaleta 1 +Mark_Richt 3 Wilton_Gregory 1 +Martha_Smith 1 Ray_Sherman 1 +Martin_ONeill 1 Mike_Weir 11 +Mary-Kate_Olsen 1 Sim_Yong 1 +Matt_Siebrandt 1 Rodney_Dangerfield 1 +Maurice_Cheeks 1 Steve_Austin 1 +Mehmet_Okur 1 Randy_Travis 1 +Michael_McNeely 1 Sean_Combs 1 +Michael_Milton 1 Wilfredo_Moreno 1 +Michael_Powell 1 Scott_Weiland 1 +Miguel_Rosseto 1 Pascal_Rheaume 1 +Miguel_Rosseto 1 Ricky_Ponting 2 +Mike_Flanagan 1 Mohammed_Baqir_al-Hakim 2 +Mike_Tice 1 Patti_Smith 1 +Mohammed_Baqir_al-Hakim 2 Tatyana_Tomashova 1 +Mufti_Mohammad_Syed 1 Raoul_Ruiz 4 +Nancy_Demme 1 Roger_Mahony 1 +Nancy_Reagan 1 Rafael_Bielsa 1 +Natasha_McElhone 2 Robert_Nardelli 1 +Nathirah_Hussein 1 Susan_Whelan 1 +Neil_Moritz 1 Roy_Williams 2 +Neil_Moritz 1 Xanana_Gusmao 5 +Nelson_Shanks 1 Pedro_Martinez 1 +Nelson_Shanks 1 Ren_Qingjin 1 +Noor_Mohammed 1 Scott_Weiland 1 +Oswald_Gruebel 1 Richard_Penniman 1 +Parthiv_Patel 1 Tonya_Payne 1 +Pauley_Perrette 1 Roy_Williams 4 +Peter_Arnett 2 Rodney_Dangerfield 1 +Peter_Greenspun 1 Rod_Thorn 1 +Peter_Harvey 1 Warren_Truss 1 +Raja_Ramani 1 Richard_Greenberg 1 +Raja_Ramani 1 Will_Young 1 +Ralph_Sampson 1 Samantha_Daniels 1 +Raoul_Ruiz 3 Tirunesh_Dibaba 1 +Ren_Qingjin 1 Wilton_Gregory 1 +Ricardo_Lopez_Murphy 2 Roberto_Robaina 1 +Ricky_Ponting 1 Ronaldo_Luis_Nazario_de_Lima 4 +Robin_Cook 1 Wilton_Gregory 1 +Roger_Federer 4 Steve_Backley 2 +Ronald_Post 1 Steve_Karsay 1 +Ryan_Nyquist 1 Winona_Ryder 22 +Sally_Clark 1 Scott_Ritter 2 +Samantha_Daniels 1 Taufik_Hidayat 3 +Sherri_Coale 1 Troy_Garity 1 +Steve_Valentine 1 Toni_Braxton 2 +Tammy_Helm 1 Tom_Smothers 1 +Tanya_Lindenmuth 1 Tora_Takagi 1 +Toni_Braxton 3 Yann_Martel 2 +Adam_Scott 1 2 +Ahmad_Masood 1 2 +Alan_Mulally 1 2 +Alexander_Rumyantsev 1 2 +Ali_Abbas 1 2 +Amanda_Bynes 1 2 +Amanda_Bynes 1 3 +Amanda_Bynes 1 4 +Amanda_Bynes 2 3 +Andrei_Mikhnevich 1 2 +Angela_Merkel 1 2 +Angela_Merkel 4 5 +Angelina_Jolie 1 15 +Angelina_Jolie 2 11 +Angelina_Jolie 8 15 +Angelina_Jolie 9 15 +Angelina_Jolie 11 20 +Anthony_LaPaglia 1 2 +Arlen_Specter 1 2 +Atal_Bihari_Vajpayee 4 6 +Barry_Alvarez 1 2 +Bill_Belichick 1 2 +Bill_Sizemore 1 2 +Boris_Yeltsin 1 2 +Britney_Spears 4 11 +Bruce_Van_De_Velde 1 2 +Camilla_Parker_Bowles 1 2 +Carolina_Kluft 1 2 +Carson_Palmer 1 3 +Carson_Palmer 2 3 +Catherine_Deneuve 1 4 +Catherine_Deneuve 2 5 +Catherine_Deneuve 3 5 +Cesar_Maia 1 2 +Charles_Moose 10 12 +Chen_Liang_Yu 1 2 +Choi_Sung-hong 1 5 +Choi_Sung-hong 2 4 +Choi_Sung-hong 3 4 +Christina_Aguilera 1 2 +Christina_Aguilera 1 4 +Christina_Aguilera 2 4 +Clara_Harris 1 5 +Clara_Harris 2 4 +Coretta_Scott_King 1 2 +Coretta_Scott_King 1 3 +Coretta_Scott_King 2 3 +Courtney_Love 1 2 +Daniel_Day-Lewis 1 3 +Daniel_Day-Lewis 2 3 +Daniela_Hantuchova 1 2 +Debra_Messing 1 2 +Dennis_Kucinich 2 7 +Dick_Latessa 1 2 +Donald_Rumsfeld 39 110 +Donald_Rumsfeld 51 83 +Donald_Rumsfeld 60 67 +Eduard_Shevardnadze 1 5 +Edward_Lu 4 6 +Edwina_Currie 1 2 +Edwina_Currie 2 4 +Elizabeth_Shue 1 2 +Ellen_Engleman 1 2 +Emma_Thompson 1 2 +Emma_Thompson 1 3 +Emma_Thompson 2 3 +Emmit_Smith 1 2 +Fayssal_Mekdad 1 3 +Fayssal_Mekdad 1 4 +Frances_Fisher 1 2 +Frank_Griswold 1 2 +Gary_Forsee 1 2 +Gloria_Macapagal_Arroyo 3 12 +Gloria_Macapagal_Arroyo 3 25 +Gloria_Macapagal_Arroyo 8 23 +Gregory_Hines 1 2 +Gus_Van_Sant 1 2 +Gus_Van_Sant 1 3 +Gwendal_Peizerat 1 2 +Gwendal_Peizerat 1 3 +Hal_Gehman 2 4 +Hannah_Stockbauer 1 2 +Harrison_Ford 1 12 +Heizo_Takenaka 1 7 +Heizo_Takenaka 3 8 +Heizo_Takenaka 3 9 +Hichiro_Naemura 1 2 +Hipolito_Mejia 1 3 +Hipolito_Mejia 1 4 +Holly_Hunter 2 6 +Holly_Hunter 3 7 +Igor_Ivanov 5 16 +Intisar_Ajouri 2 3 +Jack_Nicholson 1 3 +James_Blake 1 10 +James_Blake 4 6 +James_Blake 5 9 +James_Kelly 4 5 +James_Kopp 2 4 +James_Smith 1 2 +Jean_Chretien 2 39 +Jean_Chretien 6 40 +Jean_Chretien 8 35 +Jean_Chretien 19 28 +Jean_Chretien 28 52 +Jean_Chretien 29 37 +Jean_Chretien 34 36 +Jeffrey_Immelt 1 2 +Jennifer_Aniston 1 7 +Jennifer_Aniston 1 16 +Jennifer_Aniston 6 21 +Jerry_Falwell 1 2 +Jesse_Harris 1 3 +Jesse_Harris 2 3 +Joan_Claybrook 1 2 +Joe_Calzaghe 1 2 +Joe_Nichols 1 3 +Joe_Nichols 2 4 +Joe_Nichols 3 4 +Joerg_Haider 1 2 +John_Abizaid 1 2 +John_Abizaid 1 7 +John_Abizaid 2 3 +John_Abizaid 4 7 +John_Abizaid 4 8 +John_Abizaid 5 6 +John_Jumper 1 2 +John_Mayer 1 2 +John_Mayer 1 3 +John_Negroponte 5 26 +John_Negroponte 8 17 +John_Negroponte 28 30 +John_Reilly 1 2 +John_Rowland 1 2 +Johnny_Depp 1 2 +Jon_Corzine 1 2 +Jose_Dirceu 1 2 +Jose_Sarney 1 3 +Joseph_Estrada 1 3 +Joseph_Estrada 2 3 +Juan_Carlos_Ferrero 9 17 +Jude_Law 1 2 +Katherine_Harris 1 2 +Katherine_Harris 1 4 +Katherine_Harris 3 4 +Kelly_Clarkson 1 2 +Kiki_Vandeweghe 1 2 +Kofi_Annan 8 14 +Kofi_Annan 8 32 +Kofi_Annan 13 25 +Lance_Bass 1 4 +Lance_Bass 2 4 +Lance_Bass 2 5 +Lance_Bass 4 5 +Laurent_Gbagbo 1 2 +Lee_Soo-hyuck 1 2 +Leslie_Caldwell 2 3 +Li_Zhaoxing 2 7 +Liu_Mingkang 1 2 +Luciano_Pavarotti 1 2 +Luis_Gonzalez_Macchi 1 2 +Luis_Gonzalez_Macchi 2 4 +Luis_Gonzalez_Macchi 3 4 +Luis_Gonzalez_Macchi 3 5 +Luis_Horna 1 4 +Luis_Horna 2 4 +Luiz_Inacio_Lula_da_Silva 2 9 +Luiz_Inacio_Lula_da_Silva 2 39 +Luiz_Inacio_Lula_da_Silva 4 8 +Luiz_Inacio_Lula_da_Silva 7 36 +Luiz_Inacio_Lula_da_Silva 10 19 +Luiz_Inacio_Lula_da_Silva 11 44 +Luiz_Inacio_Lula_da_Silva 28 44 +Madeleine_Albright 1 2 +Madeleine_Albright 2 3 +Marcus_Gronholm 1 2 +Marissa_Jaret_Winokur 1 2 +Mark_Geragos 1 2 +Mark_Philippoussis 2 7 +Mark_Philippoussis 3 11 +Mark_Philippoussis 4 6 +Mark_Philippoussis 4 9 +Mark_Philippoussis 6 7 +Mark_Philippoussis 6 9 +Mark_Philippoussis 7 9 +Markus_Naslund 1 2 +Martha_Lucia_Ramirez 2 4 +Martha_Lucia_Ramirez 3 4 +Meryl_Streep 14 15 +Mesut_Yilmaz 1 2 +Michael_Schumacher 2 13 +Michael_Schumacher 2 17 +Michael_Schumacher 10 11 +Michael_Schumacher 13 18 +Michel_Temer 1 2 +Michelle_Yeoh 1 2 +Michelle_Yeoh 2 5 +Monica_Seles 1 3 +Monica_Seles 2 6 +Monica_Seles 4 5 +Moshe_Katsav 1 2 +Moshe_Katsav 2 3 +Muhammad_Saeed_al-Sahhaf 2 5 +Muhammad_Saeed_al-Sahhaf 3 4 +Muhammad_Saeed_al-Sahhaf 3 5 +Nancy_Sinatra 1 2 +Nicanor_Duarte_Frutos 3 6 +Nicanor_Duarte_Frutos 3 7 +Nicanor_Duarte_Frutos 3 11 +Patti_Labelle 2 3 +Paul_Pierce 1 2 +Paul_Shanley 2 3 +Phil_Mickelson 1 2 +Princess_Anne 1 2 +Princess_Elisabeth 1 2 +Priscilla_Owen 1 2 +Queen_Rania 2 3 +Rachel_Hunter 1 4 +Rachel_Hunter 2 3 +Rachel_Hunter 3 4 +Rafael_Ramirez 1 2 +Rafael_Ramirez 2 3 +Rafael_Ramirez 2 4 +Ralph_Lauren 1 2 +Recep_Tayyip_Erdogan 2 16 +Recep_Tayyip_Erdogan 17 19 +Reese_Witherspoon 3 4 +Richard_Armitage 2 5 +Richard_Armitage 6 7 +Richard_Armitage 6 9 +Richard_Armitage 8 9 +Richard_Gephardt 2 10 +Richard_Gephardt 3 11 +Richard_Gephardt 5 9 +Richard_Gephardt 10 11 +Richard_Norton-Taylor 1 2 +Richie_Adubato 1 2 +Robert_Bullock 1 2 +Robert_Mueller 4 5 +Robert_Mugabe 1 2 +Robert_Zoellick 2 3 +Robert_Zoellick 3 7 +Roberto_Benigni 1 2 +Roh_Moo-hyun 28 32 +Roman_Polanski 2 3 +Roman_Polanski 4 6 +Ron_Howard 1 2 +Ronald_Reagan 1 3 +Ronald_Reagan 2 3 +Rosemarie_Stack 1 2 +Roy_Jones_Jr 1 2 +Rupert_Murdoch 1 2 +Saburo_Kawabuchi 1 2 +Sam_Mendes 1 2 +Sam_Torrance 1 2 +Sam_Torrance 1 3 +Sam_Torrance 2 3 +Sergei_Ivanov 1 3 +Sergio_Vieira_De_Mello 1 2 +Sergio_Vieira_De_Mello 4 9 +Sergio_Vieira_De_Mello 7 8 +Sergio_Vieira_De_Mello 8 11 +Shannon_OBrien 1 2 +Sheila_Fraser 1 2 +Shimon_Peres 1 3 +Shimon_Peres 3 4 +Shimon_Peres 4 7 +Sourav_Ganguly 1 3 +Stanley_McChrystal 1 3 +Stanley_McChrystal 2 3 +Steve_Lavin 1 4 +Strom_Thurmond 1 3 +Surakait_Sathirathai 1 2 +Susan_Collins 1 2 +Tariq_Aziz 3 5 +Terry_Stotts 1 2 +Thaksin_Shinawatra 5 6 +Thomas_OBrien 3 9 +Tim_Chapman 1 2 +Tom_Hanks 1 7 +Tom_Hanks 4 6 +Tom_Hanks 4 9 +Tom_Hanks 5 6 +Tom_Reilly 1 2 +Tom_Reilly 2 3 +Tommy_Franks 2 4 +Tommy_Franks 4 8 +Valery_Giscard_dEstaing 1 2 +Valery_Giscard_dEstaing 1 6 +Vanessa_Williams 1 2 +Vanessa_Williams 2 3 +Victoria_Beckham 1 2 +Victoria_Beckham 1 3 +Victoria_Beckham 2 3 +Vladimir_Putin 16 21 +Vladimir_Putin 22 31 +Warren_Beatty 1 2 +Yasar_Yakis 1 4 +Yoko_Ono 1 4 +Yoko_Ono 2 4 +AJ_Lamas 1 Chris_Penn 1 +Ahmed_Qureia 1 Stanley_McChrystal 3 +Ahmet_Necdet_Sezer 1 John_Rowe 1 +Alan_Dreher 1 Emily_Mason 1 +Alan_Dreher 1 Francisco_Santos 1 +Alan_Dreher 1 Robert_Bullock 2 +Alan_Mulally 2 Tomomi_Morita 1 +Alan_Stonecipher 1 Luis_Gonzalez 1 +Alberto_Ruiz_Gallardon 1 James_Smith 2 +Alessandra_Cerna 1 John_Darby 1 +Alessandra_Cerna 1 John_Rowland 2 +Alessandra_Cerna 1 Marina_Canetti 1 +Alessandra_Cerna 1 Randy_Jackson 1 +Alessandra_Cerna 1 Richard_Jewell 1 +Alex_Cabrera 1 Fred_Huff 1 +Alex_Cabrera 1 Tommy_Franks 9 +Alex_Corretja 1 Jacky_Cheung 1 +Alexandra_Spann 1 Beth_Blough 1 +Ali_Hammoud 1 Randy_Jackson 1 +Almeida_Baptista 1 Jan_Peter_Balkenende 1 +Almeida_Baptista 1 Jason_Mewes 1 +Almeida_Baptista 1 Taoufik_Mathlouthi 1 +Amanda_Bynes 2 Grant_Rossenmeyer 1 +Amy_Cotton 1 Matt_Walters 1 +Amy_Cotton 1 Roberto_Guaterroma 1 +Andrew_Jarecki 1 Barbara_Boxer 1 +Andrew_Jarecki 1 Tessa_Jowell 1 +Andy_North 1 Eric_Hinske 1 +Andy_North 1 Marina_Kuptsova 1 +Andy_North 1 Phil_Bennett 1 +Angela_Merkel 2 Hal_Gehman 1 +Angela_Merkel 3 Anna_Jones 1 +Angela_Merkel 3 Frank_Griswold 1 +Anna_Jones 1 Ray_Romano 8 +Anthony_LaPaglia 1 Roberto_Guaterroma 1 +Arlen_Specter 1 Bill_Butler 1 +Arlen_Specter 2 Barbara_Boxer 1 +Atal_Bihari_Vajpayee 3 Ruben_Sierra 1 +Atsushi_Sato 1 Monica_Serra 1 +Baburam_Bhattari 1 Muhammad_Saeed_al-Sahhaf 4 +Barry_Alvarez 1 John_Jones 1 +Bill_Belichick 1 John_Petty 1 +Bill_Belichick 1 Phillipe_Comtois 1 +Bill_Belichick 2 Roger_Machado 1 +Bill_Byrne 1 Bill_Sizemore 1 +Bill_Curry 1 Ellen_Martin 1 +Bill_Curry 1 Richard_Langille 1 +Bill_Curry 1 Roberto_Guaterroma 1 +Billy_Beane 1 Tom_Welch 1 +Boris_Trajkovski 1 Laurent_Gbagbo 2 +Brad_Brownell 1 Hussam_Mohammed_Amin 1 +Brandon_Fails 1 Christian_Lirette 1 +Brandon_Fails 1 Ellen_Engleman 2 +Brandon_Inge 1 Eric_Lloyd 1 +Brenda_Magana 1 Nikolay_Davydenko 1 +Brian_Jordan 1 Joe_Cravens 1 +Brian_Lara 1 John_Darby 1 +Brian_Lara 1 Stanley_Nelson 1 +Calvin_Harrison 1 Luis_Gonzalez_Macchi 3 +Calvin_Harrison 1 Richard_Gephardt 9 +Calvin_Harrison 1 Suzanne_Fox 1 +Camilla_Parker_Bowles 2 Gustavo_Franco 1 +Carina_Lau_Ka-ling 1 Lin_Yi-fu 1 +Carlos_Ortega 3 Lionel_Hampton 1 +Carolina_Kluft 3 Gwen_Stefani 1 +Carson_Palmer 3 Richard_Jewell 1 +Catherine_Deneuve 4 Jade_Jagger 1 +Catriona_Le_May_Doan 1 Craig_Burley 1 +Catriona_Le_May_Doan 1 Phil_Mickelson 1 +Charlie_Deane 1 Queen_Silvia 1 +Charlotte_Church 1 Kaoru_Hasuike 1 +Charlotte_Church 1 Sourav_Ganguly 1 +Cheryl_Hines 1 Du_Qinglin 1 +Cheryl_Hines 1 Lane_Odom 1 +Chin-Hui_Tsao 1 Rosemarie_Stack 1 +Choi_Sung-hong 4 Debra_Messing 1 +Choi_Sung-hong 4 Frederique_van_der_Wal 1 +Chris_Penn 1 Phil_Bennett 1 +Chris_Penn 1 Taoufik_Mathlouthi 1 +Christian_Lirette 1 Sargis_Sargsian 1 +Christina_Aguilera 2 Joshua_Harapko 1 +Christina_Aguilera 4 Jerry_Falwell 2 +Christopher_Amolsch 1 Steve_Pagliuca 1 +Christopher_Matero 1 Valery_Giscard_dEstaing 1 +Claudine_Farrell 1 Mike_Miller 2 +Clive_Woodward 1 Douglas_Faneuil 1 +Clive_Woodward 1 Jose_Sarney 2 +Coretta_Scott_King 1 Robert_Torricelli 1 +Cori_Enghusen 1 Ernie_Stewart 1 +Courtney_Love 1 Jennifer_Aniston 3 +Courtney_Love 1 Ruben_Sierra 1 +Craig_Burley 1 Stanley_McChrystal 3 +Daniel_Coats 1 Kathryn_Grayson 1 +Daniel_Day-Lewis 3 Scott_Yates 1 +Daniela_Hantuchova 1 Jan_Peter_Balkenende 1 +Danny_Green 1 Rodrigo_Rato 1 +Darin_Erstad 1 Steve_Fehr 1 +Daryl_Smith 1 Reese_Witherspoon 4 +David_Hanson 1 Richard_Norton-Taylor 1 +David_Hilt 1 Hipolito_Mejia 3 +Dennis_Franchione 1 Hugh_Miller 1 +Dennis_Kucinich 3 Sam_Torrance 2 +Dennis_Kucinich 4 Radovan_Karadzic 1 +Dennis_Kucinich 6 Iain_Anderson 1 +Derek_Bond 1 John_Jumper 1 +Diane_Ladd 1 Trevor_Watson 1 +Dick_Latessa 1 John_Burnett 1 +Dick_Latessa 2 Martha_Beatriz_Roque 1 +Didier_Defago 1 Jerry_Falwell 2 +Don_Hewitt 1 Guennadi_Chipouline 1 +Donald_Carty 1 Ernie_Stewart 1 +Donald_Carty 1 Jeffrey_Immelt 2 +Donald_Carty 1 Vladimir_Putin 34 +Du_Qinglin 1 Meryl_Streep 7 +Dustan_Mohr 1 Edward_Flynn 1 +Dustan_Mohr 1 Enrik_Vendt 1 +Dustan_Mohr 1 Nikolay_Davydenko 1 +Dustan_Mohr 1 Sourav_Ganguly 5 +Dustan_Mohr 1 Theo_Angelopoulos 1 +E_Clay_Shaw 1 Jerry_Jones 1 +Eduard_Shevardnadze 3 Martha_Lucia_Ramirez 1 +Edward_Flynn 1 Madeleine_Albright 2 +Edward_Lu 1 Robert_Zoellick 7 +Edwina_Currie 4 Eric_Hinske 2 +Elena_de_Chavez 1 Jack_Goodman 1 +Elizabeth_Shue 1 Odai_Hussein 1 +Elizabeth_Shue 2 Princess_Hisako 1 +Elizabeth_Shue 2 Terry_Stotts 1 +Ellen_Engleman 1 Guennadi_Chipouline 1 +Ellen_Saracini 1 Ray_Romano 6 +Elsa_Zylberstein 2 Matt_Braker 1 +Elvis_Stojko 1 Jerry_Falwell 1 +Elvis_Stojko 1 Robert_Mueller 5 +Emma_Thompson 3 Nathan_Doudney 1 +Emmit_Smith 2 Rafael_Ramirez 4 +Enrik_Vendt 1 Lane_Odom 1 +Enrique_Oliu 1 Markus_Naslund 1 +Enrique_Oliu 1 William_McDonough 1 +Eric_Hinske 1 Juan_Antonio_Samaranch 1 +Eric_Lloyd 1 Jessica_Alba 2 +Eric_Lloyd 1 Valery_Giscard_dEstaing 1 +Ernie_Harwell 1 Queen_Beatrix 3 +Felipe_Perez_Roque 2 MC_Hammer 1 +Frank_Griswold 1 Yana_Klochkova 1 +Frank_Zappa 1 Markus_Naslund 1 +Frederique_van_der_Wal 1 Pedro_Mahecha 1 +Frederique_van_der_Wal 1 Raja_Zafar-ul-Haq 1 +Gary_Leon_Ridgway 1 Robert_Wagner 1 +Gary_Leon_Ridgway 1 Scott_Yates 1 +Gary_Leon_Ridgway 1 Yana_Klochkova 1 +Gene_Keady 1 Mark_Sacco 1 +George_Gregan 1 John_Negroponte 27 +Glen_DaSilva 1 Rachel_Hunter 4 +Glen_DaSilva 1 Robert_Mugabe 2 +Glen_Sather 1 Leslie_Caldwell 3 +Glen_Sather 1 Trevor_Watson 1 +Goran_Zivkovic 1 Lee_Soo-hyuck 2 +Goran_Zivkovic 1 Martha_Sahagun_de_Fox 1 +Goran_Zivkovic 1 Shane_Reynolds 1 +Grant_Rossenmeyer 1 Trevor_Watson 1 +Gregory_Hines 1 Marcus_Gronholm 2 +Guennadi_Chipouline 1 Tom_Lantos 1 +Guillermo_Ortiz 1 Nestor_Santillan 1 +Guillermo_Ortiz 2 Strom_Thurmond 2 +Gus_Van_Sant 3 Natalie_Stewart 1 +Gwen_Stefani 1 Hugh_Miller 1 +Gwendal_Peizerat 3 Turner_Stevenson 1 +Hal_Gehman 1 Scott_Yates 1 +Hamad_Bin_Isa_al-Khalifa 1 Marta_Dominguz 1 +Harrison_Ford 2 Marcus_Gronholm 1 +Harvey_Wachsman 1 Narayan_Singh_Pun 1 +Heather_Locklear 1 John_Mayer 3 +Heather_Locklear 1 Pablo_Khulental 1 +Hichiro_Naemura 2 Richard_Jewell 1 +Hipolito_Mejia 3 Jim_Ahern 1 +Holly_Hunter 3 Joshua_Perper 1 +Hugh_Hefner 1 Sam_Mendes 1 +Iain_Anderson 1 James_Kopp 2 +Iain_Anderson 1 Recep_Tayyip_Erdogan 12 +Ilan_Goldfajn 1 Jennifer_Aniston 17 +Ilie_Nastase 1 Michael_Lopez-Alegria 1 +Intisar_Ajouri 3 Lin_Yi-fu 1 +Itamar_Franco 1 Jessica_Alba 1 +Jack_Goodman 1 Kirsten_Dunst 1 +Jack_Goodman 1 Saburo_Kawabuchi 2 +Jack_Nicholson 1 Pat_Wharton 1 +Jack_Valenti 1 Shinzo_Abe 1 +Jake_Plummer 1 Jose_Dirceu 2 +Jakob_Kellenberger 1 Lara_Logan 1 +Jakob_Kellenberger 1 Pedro_Mahecha 1 +James_Maguire 1 Ronald_Reagan 1 +James_Smith 2 Kyra_Sedgwick 1 +Jason_Statham 1 Kwame_Kilpatrick 1 +Jawad_Boulus 1 Narayan_Singh_Pun 1 +Jean-Rene_Fourtou 1 Roh_Moo-hyun 28 +Jeannette_Biedermann 1 Kenny_Brack 1 +Jennifer_Granholm 1 Trudi_Lacey 1 +Jerelle_Kraus 1 Warren_Beatty 1 +Jesus_Cardenal 1 Pablo_Khulental 1 +Jim_Hendry 1 Larry_Beinfest 1 +Joan_Claybrook 2 Pablo_Latras 1 +Joaquin_Sanchez 1 Richard_Armitage 1 +Joe_Cocker 1 Martha_Beatriz_Roque 1 +Joe_Cocker 1 Randy_Jackson 1 +Joe_Cravens 1 Sam_Mendes 1 +John_Barnett 1 Kathryn_Grayson 1 +John_Darby 1 Masum_Turker 3 +John_Ferguson 1 Paul_Schrader 1 +John_Paul_DeJoria 1 Paul_Pierce 1 +John_Reilly 2 Princess_Hisako 1 +John_Rowland 2 Steve_Kerr 1 +Johnny_Depp 1 Leo_Mullin 1 +Johnny_Depp 1 Trudi_Lacey 1 +Johnny_Htu 1 Roger_Machado 1 +Jose_Dirceu 1 Robert_Torricelli 3 +Juan_Antonio_Samaranch 1 Stefan_Koubek 1 +Juergen_Braehmer 1 Shawn_Bradley 1 +Juergen_Braehmer 1 Travis_Rudolph 1 +Kathryn_Grayson 1 Queen_Beatrix 4 +Kathryn_Grayson 1 Ruben_Sierra 1 +Kathryn_Grayson 1 Shannon_OBrien 2 +Katie_Holmes 1 Park_Jung_Sung 1 +Kenny_Brack 1 Lin_Yi-fu 1 +Kwame_Kilpatrick 1 Shannon_OBrien 2 +Kyra_Sedgwick 1 Reese_Witherspoon 3 +Kyra_Sedgwick 1 Richard_Hellfant 1 +Kyra_Sedgwick 1 Wilbert_Elki_Meza_Majino 1 +Lance_Bass 3 Queen_Rania 5 +Lara_Logan 1 Turner_Stevenson 1 +Laszlo_Kovacs 1 Luke_Ridnour 1 +Laszlo_Kovacs 1 Olesya_Bonabarenko 2 +Lee_Byung-woong 1 William_Hochul 2 +Lee_Soo-hyuck 1 Robert_Bullock 1 +Leslie_Caldwell 1 Shane_Reynolds 1 +Leslie_Caldwell 2 Turner_Stevenson 1 +Li_Zhaoxing 4 Susan_Sarandon 4 +Linda_Mason 1 Rupert_Murdoch 2 +Liu_Mingkang 2 Pedro_Pauleta 1 +Liu_Ye 1 Ruben_Sierra 1 +Liu_Ye 1 Steve_Alford 1 +Loretta_Lynn_Harper 1 Michel_Temer 2 +Luis_Gonzalez 1 Warren_Beatty 1 +Madeleine_Albright 2 Martha_Lucia_Ramirez 2 +Maha_Habib 1 Princess_Elisabeth 1 +Marc_Anthony 1 Michael_Pfleger 1 +Marcus_Gronholm 1 Teresa_Graves 1 +Marcus_Gronholm 1 Tommy_Franks 8 +Marina_Canetti 1 Saied_Hadi_al_Mudarissi 1 +Mario_Alfaro-Lopez 1 Richard_Armitage 6 +Mario_Alfaro-Lopez 1 Tom_Lantos 1 +Mark_Dacey 1 Turner_Stevenson 1 +Mark_Dacey 2 Steve_Cox 1 +Mark_Kelly 1 Mark_Lazarus 1 +Mark_Lazarus 1 Ronnie_Jagday 1 +Mark_Sacco 1 Tono_Suratman 1 +Markus_Naslund 1 Robert_Wagner 1 +Martin_Kristof 1 Taia_Balk 1 +Matt_Anderson 1 Seth_Gorney 1 +Matt_Braker 1 Surakait_Sathirathai 1 +Matt_Roney 1 Wilbert_Elki_Meza_Majino 1 +Michael_Bolton 1 Robert_Bullock 2 +Michael_Pfleger 1 Robert_Ehrlich 2 +Michael_Schumacher 17 Natalie_Imbruglia 1 +Michael_Shane_Jolly 1 Trudi_Lacey 1 +Michel_Temer 1 Paul_Shanley 1 +Michelle_Yeoh 5 Saburo_Kawabuchi 1 +Mike_Helton 1 Octavio_Lara 1 +Mike_Miller 2 Toby_Keith 1 +Monica_Seles 2 Queen_Beatrix 4 +Nancy_Sinatra 1 Sargis_Sargsian 1 +Pablo_Khulental 1 Rosemarie_Stack 2 +Park_Jung_Sung 1 Robert_Bullock 2 +Paul_Pierce 2 Robert_Bullock 2 +Pedro_Pauleta 1 Robert_Tyrrell 1 +Pedro_Pauleta 1 Tommy_Franks 14 +Phil_Morris 1 Priscilla_Owen 2 +Phil_Morris 1 Roberto_Guaterroma 1 +Phil_Morris 1 Yuri_Luzhkov 1 +Queen_Rania 3 Ronnie_Jagday 1 +Radovan_Karadzic 1 Richard_Langille 1 +Randy_Jackson 1 Steve_Alford 1 +Reese_Witherspoon 1 Sam_Torrance 1 +Reese_Witherspoon 2 Toni_Jennings 1 +Rina_Lazo 1 Ronald_Reagan 1 +Robert_Ehrlich 2 Ron_Lantz 1 +Robert_Mugabe 1 Tessa_Jowell 1 +Robert_Torricelli 3 Taoufik_Mathlouthi 1 +Robert_Tyrrell 1 Yuri_Luzhkov 1 +Robert_Wagner 1 Tommy_Franks 11 +Robin_Wright_Penn 1 Sam_Torrance 3 +Robin_Wright_Penn 1 Yoko_Ono 6 +Rogelio_Ramos 1 Ryan_Newman 1 +Roman_Polanski 4 Toby_Keith 1 +Ronnie_Jagday 1 Sidney_Kimmel 1 +Scott_Yates 1 Steve_Cox 1 +Sean_Patrick_Thomas 1 William_Hochul 1 +Sharon_Osbourne 2 Shimon_Peres 6 +Stefan_Koubek 1 Steve_Alford 1 +Teri_Garr 1 Yana_Klochkova 1 +Warren_Beatty 2 William_McDonough 1 +Aleksander_Kwasniewski 1 2 +Aleksander_Kwasniewski 2 4 +Aleksander_Kwasniewski 3 4 +Ali_Naimi 1 2 +Ali_Naimi 1 5 +Ali_Naimi 2 8 +Amanda_Beard 1 2 +Andrew_Cuomo 1 2 +Arnold_Schwarzenegger 8 31 +Arnold_Schwarzenegger 9 39 +Arnold_Schwarzenegger 11 41 +Arnold_Schwarzenegger 12 37 +Arnold_Schwarzenegger 13 39 +Arnold_Schwarzenegger 22 33 +Arnold_Schwarzenegger 28 41 +Bernard_Lord 1 2 +Beth_Jones 1 2 +Branko_Crvenkovski 2 3 +Brigitte_Boisselier 1 2 +Bruce_Springsteen 1 2 +Bruce_Springsteen 1 3 +Bruce_Springsteen 2 3 +Bruce_Springsteen 2 4 +Calista_Flockhart 1 5 +Calista_Flockhart 4 6 +Cameron_Diaz 1 3 +Cameron_Diaz 3 6 +Carol_Moseley_Braun 1 2 +Carrie-Anne_Moss 1 3 +Carrie-Anne_Moss 1 5 +Carrie-Anne_Moss 2 5 +Carrie-Anne_Moss 3 5 +Carrie-Anne_Moss 4 5 +Charlton_Heston 1 5 +Charlton_Heston 2 5 +Charlton_Heston 4 6 +Chris_Bell 1 2 +Chung_Mong-hun 1 2 +Ciro_Gomes 1 2 +Ciro_Gomes 2 5 +Ciro_Gomes 3 5 +Colin_Montgomerie 1 4 +Colin_Montgomerie 2 3 +Colin_Montgomerie 2 5 +Colin_Montgomerie 4 5 +David_Trimble 1 4 +David_Trimble 1 5 +David_Trimble 3 4 +David_Wolf 1 2 +Demetrius_Ferraciu 1 2 +Dennis_Powell 1 2 +Desiree_Lemosi 1 2 +Dolma_Tsering 1 2 +Doug_Melvin 1 2 +Edward_Norton 1 2 +Emanuel_Ginobili 2 3 +Eric_Clapton 1 2 +Fabiola_Zuluaga 1 2 +Faye_Dunaway 1 2 +Ferenc_Madl 1 2 +Fernando_Gonzalez 1 5 +Fernando_Gonzalez 1 8 +Fernando_Gonzalez 2 6 +Filippo_Inzaghi 1 2 +Filippo_Inzaghi 1 3 +Filippo_Inzaghi 2 3 +Francis_George 1 2 +Franz_Beckenbauer 1 2 +Franz_Muentefering 1 2 +Franz_Muentefering 3 4 +Gary_Bergeron 1 2 +George_Foreman 1 2 +George_Pataki 1 3 +George_Pataki 3 4 +George_Pataki 3 5 +George_Roy_Hill 1 2 +Gonzalo_Sanchez_de_Lozada 7 8 +Gonzalo_Sanchez_de_Lozada 7 10 +Gonzalo_Sanchez_de_Lozada 8 11 +Gonzalo_Sanchez_de_Lozada 8 12 +Gregory_Geoffroy 1 2 +Guillermo_Coria 7 13 +Guillermo_Coria 19 22 +Habib_Rizieq 1 3 +Habib_Rizieq 2 3 +Hamid_Karzai 1 3 +Hamid_Karzai 4 13 +Hamid_Karzai 19 22 +Hilmi_Ozkok 1 2 +Horst_Koehler 1 2 +Horst_Koehler 1 3 +Horst_Koehler 2 3 +Howard_Schultz 1 2 +Ilan_Ramon 1 2 +Javier_Solana 4 8 +Javier_Solana 6 9 +Javier_Solana 6 10 +Jay_Garner 1 3 +Jean-Claude_Braquet 1 2 +Jean-Claude_Trichet 1 2 +Jean_Charest 1 10 +Jean_Charest 2 11 +Jean_Charest 7 10 +Jean_Charest 8 13 +Jean_Charest 8 17 +Jefferson_Perez 1 2 +Jim_Edmonds 1 2 +John_Kerry 1 5 +John_Kerry 3 4 +John_Kerry 4 12 +John_Kerry 9 13 +John_Malkovich 1 2 +John_Malkovich 1 3 +John_McCallum 1 2 +John_McEnroe 1 2 +John_Rosa 1 3 +Johnny_Unitas 1 2 +Jong_Wook_Lee 3 4 +Jose_Maria_Aznar 1 13 +Jose_Maria_Aznar 6 13 +Jose_Maria_Aznar 6 20 +Jose_Maria_Aznar 15 23 +Jose_Maria_Aznar 17 22 +Julianne_Moore 4 9 +Julianne_Moore 6 16 +Julianne_Moore 6 19 +Julianne_Moore 7 18 +Julianne_Moore 12 15 +Julio_Iglesias_Jr 1 2 +Justin_Timberlake 1 2 +Justin_Timberlake 1 3 +Justin_Timberlake 5 6 +Justin_Timberlake 6 8 +Justine_Pasek 2 5 +Justine_Pasek 2 6 +Justine_Pasek 6 7 +Kathleen_Glynn 1 2 +Kim_Dae-jung 1 3 +Kim_Dae-jung 1 5 +Kim_Dae-jung 2 3 +Kim_Dae-jung 4 6 +Kirk_Johnson 1 2 +Kirk_Johnson 1 3 +Kirk_Johnson 2 3 +Kosuke_Kitajima 1 2 +Kurt_Russell 1 2 +Larry_Coker 1 4 +Larry_Coker 2 4 +Larry_Ellison 1 2 +Larry_Ellison 1 3 +Lawrence_MacAulay 1 2 +LeBron_James 1 4 +LeBron_James 1 5 +LeBron_James 3 4 +LeBron_James 4 5 +Lea_Fastow 1 2 +Lee_Hoi-chang 1 2 +Lee_Hoi-chang 2 4 +Li_Peng 2 6 +Li_Peng 4 7 +Li_Peng 4 8 +Li_Peng 4 9 +Lim_Dong-won 1 2 +Lindsay_Benko 1 2 +Lou_Piniella 1 3 +Lou_Piniella 2 3 +Lucio_Gutierrez 1 2 +Lucio_Gutierrez 3 10 +Lucio_Gutierrez 6 7 +Lucio_Gutierrez 6 8 +Lucio_Gutierrez 8 13 +Luis_Figo 2 3 +Luis_Figo 2 4 +Luke_Walton 1 2 +Marat_Safin 1 2 +Marat_Safin 1 3 +Marat_Safin 2 3 +Mariana_Pollack 1 3 +Mariana_Pollack 2 3 +Martin_McGuinness 3 4 +Masahiko_Nagasawa 1 2 +Michael_Chiklis 1 4 +Michael_Chiklis 1 5 +Michael_Chiklis 2 3 +Michael_Chiklis 2 4 +Michael_Chiklis 3 5 +Michael_Chiklis 4 5 +Michael_Douglas 1 2 +Michael_Douglas 1 3 +Michael_Douglas 2 4 +Michael_Douglas 3 6 +Michael_Douglas 4 5 +Michael_Kostelnik 1 2 +Michael_Leavitt 1 2 +Michelle_Branch 1 2 +Michelle_Kwan 1 4 +Michelle_Kwan 4 5 +Michelle_Kwan 6 8 +Mike_Montgomery 1 2 +Mike_Tyson 1 3 +Mikhail_Wehbe 1 3 +Minnie_Driver 1 2 +Miyako_Miyazaki 1 2 +Muammar_Gaddafi 1 2 +Munir_Akram 1 2 +Nadia_Petrova 1 2 +Nadia_Petrova 1 3 +Nadia_Petrova 1 4 +Nadia_Petrova 2 3 +Nadia_Petrova 3 4 +Nadia_Petrova 4 5 +Nancy_Pelosi 1 7 +Nancy_Pelosi 2 9 +Nancy_Pelosi 4 15 +Nancy_Pelosi 12 14 +Nia_Vardalos 1 3 +Nia_Vardalos 3 4 +Nia_Vardalos 3 5 +Nicholas_Byron 1 2 +Nikki_Reed 1 2 +Olivia_Newton-John 1 2 +Paradorn_Srichaphan 2 4 +Paradorn_Srichaphan 4 5 +Pascal_Quignard 1 3 +Pascal_Quignard 2 3 +Patrice_Chereau 1 2 +Pedro_Malan 1 5 +Pedro_Malan 3 5 +Peter_Harrison 1 2 +Rainer_Schuettler 1 4 +Rainer_Schuettler 2 3 +Rainer_Schuettler 2 4 +Raymond_Odierno 1 2 +Rebecca_Romijn-Stamos 1 2 +Rebecca_Romijn-Stamos 1 3 +Rebecca_Romijn-Stamos 1 4 +Rebecca_Romijn-Stamos 3 4 +Richard_Krajicek 1 3 +Richard_Krajicek 2 3 +Rick_Santorum 1 2 +Rick_Santorum 1 3 +Rick_Santorum 2 3 +Rick_Stansbury 2 3 +Rob_Marshall 1 3 +Rob_Marshall 1 6 +Rob_Marshall 2 4 +Robert_Blackwill 1 2 +Robert_Duvall 2 8 +Robert_Horan 1 2 +Robert_Redford 3 4 +Robert_Redford 7 8 +Robinson_Stevenin 1 2 +Roger_Clemens 1 2 +Sadie_Frost 1 2 +Saparmurat_Niyazov 1 2 +Sarah_Michelle_Gellar 1 2 +Sarah_Michelle_Gellar 1 3 +Sarah_Michelle_Gellar 2 3 +Serena_Williams 1 41 +Serena_Williams 3 32 +Serena_Williams 14 40 +Serena_Williams 41 47 +Sergey_Lavrov 1 3 +Sergey_Lavrov 1 6 +Sergey_Lavrov 2 4 +Sergey_Lavrov 3 5 +Sergey_Lavrov 3 7 +Sergey_Lavrov 4 8 +Shane_Mosley 1 2 +Sheila_Copps 2 3 +Sheila_Copps 3 4 +Shia_LaBeouf 1 2 +Steve_Nash 1 3 +Steve_Nash 1 4 +Steve_Nash 2 4 +Steve_Nash 2 5 +Steve_Spurrier 1 2 +Steve_Waugh 1 2 +Susilo_Bambang_Yudhoyono 1 2 +Susilo_Bambang_Yudhoyono 1 3 +Susilo_Bambang_Yudhoyono 1 4 +Susilo_Bambang_Yudhoyono 3 4 +Suzanne_Gaudet 1 2 +Thomas_Bjorn 1 2 +Tim_Conway 1 2 +Tim_Conway 1 3 +Tim_Robbins 1 5 +Tim_Robbins 2 4 +Tim_Robbins 3 5 +Tom_Craddick 1 3 +Tom_Craddick 2 4 +Tom_Craddick 3 4 +Tracee_Ellis_Ross 1 2 +Venus_Williams 2 9 +Vivica_Fox 1 2 +William_Ford_Jr 1 2 +William_Ford_Jr 2 6 +William_Ford_Jr 5 6 +William_Rehnquist 1 2 +Yossi_Beilin 1 2 +Aaron_Eckhart 1 Akiko_Morigami 1 +Aaron_Eckhart 1 AnFernce_Negron 1 +Aaron_Eckhart 1 Sadie_Frost 1 +Abdel_Aziz_Al-Hakim 1 Joe_Darrell 1 +Abdullah_al-Attiyah 1 Rachel_Wadsworth 1 +Abdullah_al-Attiyah 2 John_Lisowski 1 +Abraham_Foxman 1 Doug_Melvin 2 +Abraham_Foxman 1 Nikki_Reed 1 +Adam_Rich 1 Jean-Claude_Trichet 1 +Adam_Rich 1 John_Goold 1 +Adam_Rich 1 Lisa_Girman 1 +Adam_Rich 1 Roger_Clemens 1 +Adam_Rich 1 Suzanne_Somers 1 +Adrian_Fernandez 1 Nicolas_Massu 1 +Aiysha_Smith 1 Yossi_Beilin 1 +Akiko_Morigami 1 Shane_Mosley 1 +Alastair_Johnston 1 Aleksander_Kwasniewski 4 +Alastair_Johnston 1 Bill_Duffey 1 +Albert_Brooks 1 Robinson_Stevenin 1 +Albert_Brooks 1 Sheila_Copps 3 +Aleksander_Kwasniewski 1 Guangdong_Ou_Guangyuan 1 +Aleksander_Kwasniewski 2 George_Plimpton 1 +Aleksander_Kwasniewski 2 Juan_Jose_Lucas 1 +Aleksander_Kwasniewski 3 Billy_Crawford 1 +Aleksander_Voloshin 1 Kristen_Rivera 1 +Alessandro_Nesta 1 Paul_Kelleher 1 +Alfredo_di_Stefano 1 Chris_Moore 1 +Alfredo_di_Stefano 1 Maryn_McKenna 1 +Alfredo_di_Stefano 1 Shamai_Leibowitz 1 +Ali_Naimi 5 Morris_Dees 1 +Aline_Chretien 1 Guillermo_Coria 17 +Amy_Yasbeck 1 Chawki_Armali 1 +AnFernce_Negron 1 Kristen_Rivera 1 +Andrew_Cuomo 2 Todd_Petit 1 +Anil_Ramsook 1 Takeo_Hiranuma 1 +Annie_Chaplin 1 Charles_Tannok 1 +Antanas_Valionis 1 William_Rehnquist 1 +Avril_Lavigne 1 Shia_LaBeouf 2 +Barbara_Becker 1 Chris_Noth 1 +Barbara_Becker 1 Franz_Beckenbauer 2 +Barbora_Strycova 1 Christiane_Wulff 1 +Barry_Diller 1 Meles_Zenawi 1 +Bartosz_Kizierowski 1 Faye_Wong 1 +Ben_Wallace 1 Nia_Vardalos 5 +Bernard_Lord 1 Joxel_Garcia 1 +Bernard_Lord 2 Bing_Crosby 1 +Beyonce_Knowles 1 Maurice_Papon 1 +Bill_Duffey 1 Jong_Wook_Lee 1 +Bill_Guerin 1 Julie_Goodenough 1 +Bill_Herrion 1 Fernando_Valenzuela 1 +Bill_Herrion 1 Johnny_Unitas 2 +Bill_Richardson 1 George_McCloud 1 +Billy_Andrade 1 Kellie_Coffey 1 +Billy_Crawford 1 Jim_Edmonds 2 +Billy_Edelin 1 Himmler_Rebu 1 +Bing_Crosby 1 Stephen_Webster 1 +Bixente_LIzarazu 1 Chris_Bell 2 +Bixente_LIzarazu 1 Todd_Wike 1 +Blythe_Danner 2 Hank_McKinnell 1 +Bob_Eskridge 1 Marco_Pantani 1 +Bob_Hartley 1 Lou_Piniella 3 +Bob_Krueger 1 Gordana_Grubin 1 +Bob_Krueger 1 Mariana_Pollack 1 +Bob_Sulkin 1 Branko_Crvenkovski 3 +Bobby_Kielty 1 Robert_Horan 2 +Brajesh_Mishra 1 Mark_Podlesny 1 +Brandon_Knight 1 Claudia_Cardinale 1 +Brandon_Knight 1 Phil_Donahue 1 +Brandon_Lloyd 1 Cha_Yung-gu 1 +Brandon_Lloyd 1 James_Coburn 1 +Brian_Schneider 1 Michael_Rolinee 1 +Bruce_Springsteen 4 Ion_Tiriac 1 +Cameron_Diaz 2 Shia_LaBeouf 1 +Cameron_Diaz 5 Nadia_Petrova 2 +Carla_Gugino 1 Kelly_Osbourne 1 +Carla_Gugino 1 Kenny_Chesney 1 +Carla_Gugino 1 Michael_Douglas 3 +Carla_Gugino 1 Miguel_Aldana_Ibarra 1 +Carlos_Fasciolo 1 Debbie_Allen 1 +Carol_Williams 1 Gorden_Tallis 1 +Carrie-Anne_Moss 2 Horacio_Julio_Pina 1 +Carrie-Anne_Moss 4 Michael_Goldrich 1 +Casey_Crowder 1 Keiko_Sofia_Fujimori 1 +Casey_Crowder 1 Phoenix_Chang 1 +Cha_Yung-gu 1 Takeo_Hiranuma 1 +Charlene_Barshefsky 1 James_Coburn 1 +Charles_Tannok 1 Larry_Coker 4 +Charlie_Hunnam 1 Hestrie_Cloette 1 +Charlie_Hunnam 1 Veronica_Lake 1 +Charlton_Heston 4 Desiree_Lemosi 1 +Chawki_Armali 1 Shoshannah_Stern 1 +Chris_Bell 2 Pharrell_Williams 1 +Chris_Forsyth 1 Nikki_Reed 2 +Chris_Neil 1 John_Kerry 14 +Chris_Noth 1 Takeshi_Kitano 1 +Christine_Arron 1 Robert_Duvall 5 +Christine_Arron 1 Steve_McManaman 1 +Chuck_Hagel 1 Jeff_Roehm 1 +Chuck_Hagel 1 Margaret_Hasley 1 +Chuck_Hagel 1 Sebastian_Porto 1 +Cliff_Ellis 1 John_Lisowski 1 +Clifford_Etienne 1 Pharrell_Williams 1 +Clifford_Etienne 1 Ray_Lewis 1 +Clive_Lloyd 1 Harland_Braun 1 +Colin_Montgomerie 1 Horacio_Julio_Pina 1 +Daniel_Montenegro 1 Gustavo_Cisneros 1 +Daniel_Montenegro 1 Nadia_Petrova 4 +Daniel_Montenegro 1 Omar_Khan_Sharif 1 +Daniel_Montenegro 1 Rick_Santorum 1 +Danny_Glover 1 Morris_Dees 1 +Danny_Glover 1 Patrice_Chereau 2 +David_Trimble 4 John_Elway 1 +David_Wolf 2 Emanuel_Ginobili 4 +David_Wolf 2 Kathleen_Glynn 2 +David_Wolf 2 Venus_Williams 6 +Debbie_Allen 1 Sebastian_Cuattrin 1 +Demetrius_Ferraciu 2 Edward_Seaga 1 +Denise_Johnson 1 Nikki_Cascone 1 +Denise_Johnson 2 Emanuel_Ginobili 3 +Denise_Johnson 2 Filippo_Inzaghi 2 +Dennis_Powell 1 Horst_Koehler 3 +Dennis_Powell 1 Michael_Chiklis 2 +Desiree_Lemosi 2 Dion_Glover 1 +Dion_Glover 1 Michael_Weiss 1 +Doug_Melvin 1 Gustavo_Cisneros 1 +Doug_Melvin 3 Emma_Nicholson 1 +Doug_Racine 1 Marisol_Breton 1 +Ed_Book 1 Nur_Jaafar 1 +Ed_Case 1 Sadam_Hassan 1 +Edward_Seaga 1 George_Roy_Hill 1 +Edward_Seaga 1 Sarah_Price 1 +Eileen_Spina 1 Jeff_Bzdelik 1 +Ekke_Hard_Forberg 1 Nikki_Reed 1 +Elena_Bereznaya 1 Lene_Espersen 1 +Elena_Bereznaya 1 Mario_Lobo_Zagallo 1 +Elena_Bereznaya 1 Rick_Stansbury 2 +Elisha_Cuthbert 1 Mariana_Pollack 2 +Eliza_Manningham-Buller 1 Lawrence_MacAulay 2 +Eric_Clapton 2 Niall_Connolly 1 +Eric_Taino 1 Michael_Goldrich 1 +Farida_Ragoonanan 1 Jorge_Enrique_Jimenez 1 +Felix_Trinidad 1 Joe_Darrell 1 +Felix_Trinidad 1 Justine_Pasek 4 +Fernando_Valenzuela 1 Luis_Figo 3 +Fernando_Valenzuela 1 Nicolas_Macrozonaris 1 +Filippo_Inzaghi 1 Mohammed_Al_Hindi 1 +Filippo_Inzaghi 1 Shanna_Zolman 1 +Francis_Ricciardone 1 Sven_Goran_Eriksson 1 +Franz_Beckenbauer 1 Gerald_Ford 1 +Franz_Beckenbauer 2 Rick_Bragg 1 +George_Foreman 2 Ralph_Goodale 1 +George_Pataki 1 James_Dingemans 1 +George_Pataki 2 Masahiko_Nagasawa 2 +George_Plimpton 1 Ximena_Bohorquez 1 +Gerald_Ford 1 Kurt_Budke 1 +Gerald_Ford 1 Li_Peng 1 +Giselle_Estefania_Tavarelli 1 John_Sununu 1 +Giulio_Andreotti 1 Kurt_Russell 2 +Giulio_Andreotti 1 Michael_Denzel 1 +Gonzalo_Sanchez_de_Lozada 3 Shannyn_Sossamon 1 +Graciano_Rocchigiani 1 Juan_Jose_Lucas 1 +Gregory_Geoffroy 1 Sybille_Schmid 1 +Gregory_Geoffroy 2 Kosuke_Kitajima 1 +Gregory_Geoffroy 2 Wycliffe_Grousbeck 1 +Guangdong_Ou_Guangyuan 1 Li_Ruihuan 1 +Guangdong_Ou_Guangyuan 1 Nicole 1 +Gustavo_Cisneros 1 Steve_Blake 1 +Gustavo_Noboa 1 Lea_Fastow 2 +Habib_Hisham 1 Pat_Rochester 1 +Habib_Hisham 1 Paul_Wilson 1 +Habib_Rizieq 1 Olivia_Newton-John 2 +Habib_Rizieq 3 Kenny_Chesney 1 +Hal_Sutton 2 Kellie_Coffey 1 +Harland_Braun 1 Kathryn_Tucker 1 +Hassanal_Bolkiah 1 Michael_Leavitt 1 +Hestrie_Cloette 1 Nicholas_Byron 2 +Hitoshi_Tanaka 1 Ramon_Ponce_de_Leon 1 +Horacio_Julio_Pina 1 Zaini_Abdullah 1 +Huang_Suey-Sheng 1 Lucrecia_Orozco 1 +Huang_Suey-Sheng 1 Paul_Cerjan 1 +Huang_Suey-Sheng 1 Robinson_Stevenin 1 +Ian_Knop 1 Michel_Minard 1 +Ilan_Ramon 3 Tatiana_Kennedy_Schlossberg 1 +Ion_Tiriac 1 Lawrence_MacAulay 1 +Ismail_Abu_Shanab 1 Michael_Denzel 1 +Ismail_Abu_Shanab 1 Rebecca_Romijn-Stamos 1 +Jack_Welch 1 Keizo_Yamada 1 +James_Dingemans 1 Robert_Duvall 5 +James_Dingemans 1 William_Morrow 1 +Jamie_Lee_Curtis 1 William_Joppy 1 +Jamie_Martin 1 Patrick_Ewing 1 +Jan_Bjoerklund 1 Li_Ruihuan 1 +Jan_Paul_Miller 1 Paradorn_Srichaphan 8 +Jay_Garner 3 Reggie_Sanders 1 +Jeff_Bzdelik 1 Zach_Parise 1 +Jerry_Oliver 1 Rudolf_Schuster 1 +Jerry_Oliver 1 Stefaan_Declerk 1 +Jerry_Sloan 1 Julianne_Moore 9 +Jerry_Sloan 1 Keiko_Sofia_Fujimori 1 +Jerry_Sloan 1 Nikki_Cascone 1 +Jim_Anderson 1 Jose_Luis_Santiago_Vasconcelos 1 +Jim_Anderson 1 Mikhail_Wehbe 1 +Joaquim_Levy 1 Judy_Vassar 1 +Joe_Leonard 1 Kristen_Rivera 1 +John_Baldacci 1 Nona_Gaye 1 +John_Baldacci 1 Shia_LaBeouf 1 +John_Elway 1 Mahmoud_Diyab_al-Ahmed 1 +John_Kerry 9 Raja_Ibrahim 1 +John_Kerry 14 Pat_Rochester 1 +John_Lisowski 1 Michelle_Kwan 1 +John_Malkovich 2 Paradorn_Srichaphan 7 +John_Malkovich 3 Mona_Locke 1 +John_Malkovich 3 Richard_Palmer 1 +John_McCallum 1 Oliver_Phelps 1 +John_McCallum 2 Rick_Bragg 1 +John_McEnroe 2 Shane_Mosley 1 +John_Prescott 1 Will_Ofenheusle 1 +Johnnie_Lynn 1 Larry_Coker 4 +Johnnie_Lynn 1 Richard_Palmer 1 +Johnny_Unitas 1 Mark_Foley 1 +Jonathan_Karsh 1 Samantha_Ledster 1 +Joxel_Garcia 1 Lena_Olin 1 +Joy_Bryant 1 Richard_Chamberlain 1 +Julianne_Moore 4 Shia_LaBeouf 1 +Julio_Iglesias_Jr 1 Ramon_Ponce_de_Leon 1 +Justin_Timberlake 1 Suzanne_Somers 1 +Kathleen_Glynn 2 Susilo_Bambang_Yudhoyono 1 +Kathryn_Morris 1 Minnie_Driver 2 +Kathryn_Tucker 1 Stella_McCartney 1 +Katie_Couric 1 Sanjay_Gupta 1 +Keiko_Sofia_Fujimori 1 Sureyya_Ayhan 1 +Kellie_Coffey 1 Todd_Petit 1 +Kenny_Chesney 1 Steve_Coogan 1 +Kevin_Tarrant 1 Marisol_Breton 1 +Kristen_Rivera 1 Valdas_Adamkus 2 +Kurt_Russell 2 Mahmoud_Diyab_al-Ahmed 1 +Kurt_Tanabe 1 William_Jackson 1 +Kyle_McLaren 1 Sofia_Milos 1 +Kyle_McLaren 1 Will_Ofenheusle 1 +Kyle_Shewfelt 1 Larry_Coker 2 +Lawrence_Foley 1 Rachel_Wadsworth 1 +Lea_Fastow 1 Uday_Hussein 1 +Lesia_Burlak 1 Nur_Jaafar 1 +Lori_Berenson 1 Rudolf_Schuster 1 +Lucio_Angulo 1 Sarah_Price 1 +Lucio_Gutierrez 10 Pedro_Malan 5 +Margaret_Hasley 1 Michael_Denzel 1 +Mark_Redman 1 Stephen_Funk 1 +Mark_Salter 1 Shanna_Zolman 1 +Martin_Luther_King_III 1 Maryn_McKenna 1 +Mary_Bono 1 Todd_Wike 1 +Matt_LeBlanc 1 Robert_Redford 2 +Mauro_Viza 1 William_Jackson 1 +Max_Baucus 1 Paradorn_Srichaphan 5 +Max_Baucus 1 Yossi_Beilin 2 +Meles_Zenawi 1 Nawabzada_Nasrullah_Khan 1 +Melissa_Mulloy 1 Paula_Abdul 1 +Melissa_Mulloy 1 Roger_Lyons 1 +Michael_Chiklis 1 Mohammed_Al_Hindi 1 +Michael_Chiklis 1 Steve_Rush 1 +Michael_Doleac 1 Nur_Jaafar 1 +Michael_Goldrich 1 Suzanne_Somers 1 +Michael_Kahn 1 Rick_Caruso 1 +Michel_Minard 1 Suzanne_Gaudet 1 +Michelle_Bachelet 1 Sami_Al-Arian 1 +Miguel_Angel_Rodriguez 1 Sasha_Cohen 1 +Mike_Bryan 1 Shanna_Zolman 1 +Mike_Montgomery 1 Ray_Lewis 1 +Milton_Wynants 1 Stuart_Townsend 1 +Miyako_Miyazaki 2 Munir_Akram 2 +Morris_Dees 1 Shamai_Leibowitz 1 +Morris_Dees 1 Suzie_McConnell_Serio 1 +Nathalie_Gagnon 1 Richard_Reid 1 +Nicklas_Lidstrom 1 Norman_Jewison 1 +Nicklas_Lidstrom 1 Sadie_Frost 3 +Nicole_Hiltz 1 Zaini_Abdullah 1 +Nona_Gaye 1 Paul_Cerjan 1 +Oscar_Bolanos 1 Phil_Donahue 1 +Oscar_Bolanos 1 Tatiana_Kennedy_Schlossberg 1 +Pascal_Quignard 3 Patrick_Ewing 2 +Pat_Rochester 1 Phoenix_Chang 1 +Pat_Rochester 1 Will_Ofenheusle 1 +Paul_Farley 1 Platon_Lebedev 1 +Paula_Abdul 1 Robert_Vowler 1 +Pharrell_Williams 1 Tyrone_Medley 1 +Phoenix_Chang 1 Platon_Lebedev 1 +Rachel_Wadsworth 1 Richard_Palmer 1 +Raymond_Odierno 1 Richard_Reid 1 +Reggie_Sanders 1 Rick_Santorum 2 +Richard_Chamberlain 1 Steve_Patterson 1 +Richard_Ward 1 Steve_Redgrave 1 +Robert_Vowler 1 Tab_Baldwin 1 +Roy_Rogers 1 Steven_Feldman 1 +Scott_Rolen 1 William_Murabito 1 +Sofia_Milos 1 Steve_Nash 3 +Sofia_Milos 1 Will_Ofenheusle 1 +Sonja_Kesselschlager 1 Tim_Robbins 3 +Takeo_Hiranuma 1 Ty_Votaw 1 +Ted_Washington 1 Ximena_Bohorquez 1 +Ty_Votaw 1 William_Webster 1 +Adrian_McPherson 1 2 +Al_Davis 1 2 +Al_Gore 2 6 +Al_Gore 4 6 +Alan_Greenspan 1 2 +Alan_Greenspan 1 5 +Alan_Greenspan 3 4 +Alastair_Campbell 1 5 +Alastair_Campbell 2 4 +Alexander_Downer 1 2 +Alexander_Downer 1 3 +Alexander_Downer 1 4 +Alexander_Downer 3 4 +Alice_Fisher 1 2 +Alison_Lohman 1 2 +Alvaro_Silva_Calderon 1 2 +Alvaro_Silva_Calderon 1 3 +Alvaro_Uribe 7 20 +Alvaro_Uribe 8 25 +Alvaro_Uribe 12 13 +Alvaro_Uribe 20 28 +Anders_Ebbeson 2 3 +Andrew_Bunner 1 2 +Anibal_Ibarra 1 3 +Antonio_Trillanes 1 3 +Antonio_Trillanes 2 3 +Asa_Hutchinson 1 2 +Barbara_Walters 1 3 +Barbara_Walters 1 4 +Barbara_Walters 3 4 +Ben_Howland 1 2 +Ben_Howland 1 3 +Ben_Howland 3 4 +Benazir_Bhutto 1 4 +Benazir_Bhutto 2 3 +Benazir_Bhutto 2 4 +Bill_Simon 5 9 +Bill_Simon 11 15 +Billy_Sollie 1 2 +Boris_Berezovsky 1 2 +Brooke_Shields 1 2 +Bulent_Ecevit 1 4 +Bulent_Ecevit 1 5 +Bulent_Ecevit 2 6 +Bulent_Ecevit 4 6 +Candie_Kung 1 2 +Candie_Kung 1 4 +Carlo_Ancelotti 1 2 +Carlo_Ancelotti 2 3 +Carlos_Moya 8 17 +Carlos_Moya 10 18 +Carlos_Vives 1 4 +Carlos_Vives 2 4 +Carson_Daly 1 2 +Cate_Blanchett 1 2 +Cate_Blanchett 1 3 +Cate_Blanchett 1 4 +Cate_Blanchett 2 3 +Cate_Blanchett 2 4 +Chok_Tong_Goh 1 2 +Chris_Byrd 1 2 +Chris_Cooper 1 2 +Chris_Tucker 1 2 +Christine_Gregoire 1 4 +Christine_Gregoire 2 3 +Christine_Gregoire 2 4 +Christopher_Patten 1 2 +Clint_Eastwood 1 4 +Clint_Eastwood 1 6 +Constance_Marie 1 2 +Constance_Marie 1 3 +Dennis_Hastert 2 3 +Dennis_Hastert 3 6 +Dennis_Hastert 4 6 +Dennis_Hastert 5 6 +Dolly_Parton 1 2 +Doug_Duncan 1 2 +Edward_Kennedy 1 2 +Edward_Kennedy 2 3 +Edward_Said 1 2 +Elena_Bovina 1 2 +Elena_Bovina 1 3 +Elena_Bovina 2 3 +Eliane_Karp 1 3 +Eliane_Karp 2 3 +Eliane_Karp 2 4 +Eliane_Karp 3 4 +Elvis_Presley 1 2 +Erika_Harold 1 3 +Erika_Harold 2 3 +Erika_Harold 3 4 +Erika_Harold 4 5 +Ernie_Els 1 3 +Ernie_Els 1 4 +Ernie_Els 2 3 +Franco_Dragone 1 2 +Frank_Solich 1 4 +Frank_Solich 2 5 +Frank_Solich 3 4 +Gabriel_Batistuta 1 2 +Gary_Carter 1 2 +Gary_Carter 1 3 +Gary_Doer 1 2 +Gary_Doer 2 3 +George_Tenet 1 2 +George_Voinovich 1 3 +George_Voinovich 2 3 +Georgi_Parvanov 1 2 +Goldie_Hawn 1 7 +Goldie_Hawn 2 3 +Goldie_Hawn 3 7 +Goldie_Hawn 6 7 +Goran_Persson 1 2 +Gro_Harlem_Brundtland 1 2 +Guillaume_Soro 1 2 +Gwyneth_Paltrow 1 5 +Gwyneth_Paltrow 1 6 +Gwyneth_Paltrow 2 6 +Gwyneth_Paltrow 3 6 +Hee-Won_Han 1 2 +Herb_Sendek 1 3 +Herb_Sendek 3 4 +Howard_Smith 1 2 +Hugh_Grant 5 8 +Hugh_Grant 6 9 +Jack_Straw 25 28 +James_Franco 1 2 +James_Ivory 1 2 +James_Schultz 1 2 +James_Traficant 1 2 +James_Traficant 2 3 +Jan_Ullrich 1 4 +Jan_Ullrich 2 6 +Jan_Ullrich 3 6 +Jane_Pauley 1 2 +Jason_Jennings 1 2 +Javier_Weber 1 2 +Jennifer_Rodriguez 1 2 +Jeong_Se-hyun 2 6 +Jeong_Se-hyun 3 7 +Jeong_Se-hyun 6 8 +Jeong_Se-hyun 7 9 +Jeong_Se-hyun 8 9 +Jo_Dee_Messina 1 2 +Joe_Lieberman 8 11 +Joe_Lieberman 9 10 +Joe_Mantello 1 2 +John_Allen_Muhammad 2 9 +John_Allen_Muhammad 5 7 +John_Allen_Muhammad 6 10 +John_Blaney 1 2 +John_Brady 1 2 +John_Howard 5 15 +John_Howard 12 17 +Johnny_Tapia 1 2 +Johnny_Tapia 2 3 +Jorge_Valdano 1 2 +Joseph_Deiss 1 3 +Junichiro_Koizumi 1 53 +Junichiro_Koizumi 26 55 +Junichiro_Koizumi 29 45 +Kate_Capshaw 1 2 +Kathryn_Bigelow 1 2 +Kevin_Spacey 1 2 +Kevin_Spacey 1 3 +Kim_Ryong-sung 6 8 +Kim_Ryong-sung 8 11 +Klaus_Zwickel 1 2 +Kristen_Breitweiser 2 3 +Laila_Ali 2 3 +Larry_Brown 1 3 +Larry_Brown 2 4 +Larry_Brown 2 7 +Larry_Brown 4 7 +Larry_Brown 6 7 +Larry_Johnson 1 2 +Lars_Von_Trier 1 2 +Lars_Von_Trier 2 3 +Leander_Paes 1 2 +Liam_Neeson 1 3 +Liam_Neeson 2 3 +Lino_Oviedo 1 3 +Lino_Oviedo 2 3 +Ludivine_Sagnier 1 3 +Ludivine_Sagnier 3 4 +Lynne_Cheney 1 2 +Lynne_Cheney 2 3 +Maggie_Smith 1 2 +Marcelo_Salas 1 2 +Mariangel_Ruiz_Torrealba 1 2 +Marisa_Tomei 1 2 +Mary_Steenburgen 1 2 +Mel_Brooks 1 2 +Mel_Gibson 1 2 +Mian_Khursheed_Mehmood_Kasuri 2 3 +Mian_Khursheed_Mehmood_Kasuri 3 4 +Michael_Moore 1 2 +Michael_Moore 2 3 +Michel_Duclos 1 2 +Michelle_Pfeiffer 1 2 +Mireya_Moscoso 2 3 +Mireya_Moscoso 2 5 +Mireya_Moscoso 4 5 +Nabil_Shaath 1 3 +Nabil_Shaath 2 3 +Naomi_Watts 1 18 +Naomi_Watts 6 9 +Naomi_Watts 13 18 +Natalie_Cole 1 3 +Natalie_Cole 2 3 +Nathalie_Baye 1 2 +Nathalie_Baye 1 4 +Nathalie_Baye 2 4 +Nathalie_Baye 3 4 +Oleksandr_Moroz 1 2 +Oscar_Elias_Biscet 1 2 +Oswaldo_Paya 1 3 +Oswaldo_Paya 3 4 +Patricia_Clarkson 1 2 +Patricia_Clarkson 1 3 +Patricia_Clarkson 2 3 +Patricia_Clarkson 2 4 +Patrick_Roy 1 2 +Paul-Henri_Mathieu 1 2 +Paul-Henri_Mathieu 1 3 +Paul_Byrd 1 2 +Paul_Kagame 1 2 +Peter_Costello 1 2 +Peter_Greenaway 1 2 +Prince_Naruhito 1 2 +Prince_Naruhito 1 3 +Prince_Naruhito 2 3 +Princess_Masako 1 2 +Pupi_Avati 2 3 +Queen_Latifah 1 3 +Queen_Latifah 2 4 +Queen_Latifah 3 4 +Rachel_Griffiths 2 3 +Ralph_Firman 1 2 +Ralph_Klein 1 2 +Ranil_Wickremasinghe 2 3 +Rick_Pitino 1 3 +Rick_Pitino 2 4 +Rick_Pitino 3 4 +Ricky_Martin 1 2 +Rita_Moreno 1 2 +Robert_De_Niro 1 4 +Robert_De_Niro 3 6 +Robert_Kocharian 4 5 +Roberto_Marinho 2 3 +Roger_Moore 3 5 +Ron_Dittemore 1 2 +Ron_Dittemore 1 3 +Ron_Dittemore 4 6 +Rudolph_Giuliani 1 20 +Rudolph_Giuliani 2 5 +Rudolph_Giuliani 3 20 +Rudolph_Giuliani 4 17 +Rudolph_Giuliani 15 20 +Russell_Simmons 1 2 +Russell_Simmons 1 4 +Russell_Simmons 2 4 +Sean_Astin 1 3 +Sean_Astin 2 3 +Shaukat_Aziz 1 2 +Silvio_Fernandez 1 2 +Sophia_Loren 1 2 +Sophia_Loren 1 3 +Sophia_Loren 1 7 +Sophia_Loren 6 7 +Stellan_Skarsgard 1 2 +Tamara_Brooks 1 2 +Thomas_Fargo 1 2 +Thomas_Fargo 1 3 +Thomas_Fargo 2 3 +Thomas_Fargo 2 4 +Thomas_Fargo 3 4 +Tim_Allen 2 3 +Tim_Allen 3 4 +Tony_Shalhoub 1 2 +Tony_Shalhoub 1 3 +Tracy_McGrady 1 2 +Vicente_Fernandez 2 3 +Vicente_Fernandez 4 5 +Vince_Gill 1 2 +Wolfgang_Schuessel 1 4 +Wolfgang_Schuessel 3 4 +Wu_Yi 1 2 +Wu_Yi 1 3 +Wu_Yi 2 3 +Yao_Ming 2 4 +Yao_Ming 5 6 +Yao_Ming 5 8 +Yao_Ming 6 7 +Yoriko_Kawaguchi 3 5 +Yu_Shyi-kun 1 3 +Yu_Shyi-kun 1 4 +Yu_Shyi-kun 2 3 +Zhang_Wenkang 1 2 +Zinedine_Zidane 4 6 +Abdullah_Nasseef 1 Bruce_Paltrow 1 +Abdullah_Nasseef 1 Howard_Smith 2 +Abdullah_Nasseef 1 Jan_De_Bont 1 +Abdullah_Nasseef 1 Jim_Nochols 1 +Abdullah_Nasseef 1 Oleg_Romantsev 1 +Adam_Kennedy 1 Charlie_Sheen 1 +Adrian_McPherson 2 Eduardo_Chillida 1 +Al_Davis 1 Julian_Fantino 1 +Alan_Greenspan 4 Kent_McCord 1 +Alastair_Campbell 4 Li_Ka-shing 1 +Alexa_Vega 1 Tony_Shalhoub 1 +Alexander_Downer 2 Zeljko_Rebraca 1 +Alexander_Downer 3 Luis_Berrondo 1 +Alexandra_Pelosi 1 Ken_Kutaragi 1 +Alexandre_Herchcovitch 1 Karen_Clarkson 1 +Alice_Fisher 1 Gene_Sauers 1 +Alicia_Witt 1 Jorge_Moreno 1 +Alicia_Witt 1 Lino_Oviedo 1 +Allan_Wagner 1 Larry_Campbell 1 +Allan_Wagner 1 Rita_Moreno 1 +Alvaro_Silva_Calderon 2 Stepan_Demirchian 1 +Alvaro_Uribe 14 Patricia_Clarkson 1 +Ana_Sebastiao 1 Marcos_Cafu 1 +Ana_Sebastiao 1 Raza_Rabbani 1 +Andrea_De_Cruz 1 Dionne_Warwick 1 +Andrea_De_Cruz 1 Ismail_Cem 1 +Andrea_De_Cruz 1 Natalya_Sazanovich 1 +Andrea_De_Cruz 1 Yoon_Young-kwan 1 +Andres_Manuel_Lopez_Obrador 1 Jim_Parque 1 +Andres_Manuel_Lopez_Obrador 1 Norman_Mineta 1 +Andrew_Bunner 2 Oscar_Elias_Biscet 2 +Andrew_Weissmann 1 Ernie_Els 2 +Andy_Warhol 1 Edith_Masai 1 +Andy_Warhol 1 Paul-Henri_Mathieu 2 +Anibal_Ibarra 2 David_Przybyszewski 1 +Anthony_Ervin 1 Juan_Roman_Carrasco 1 +Antonio_Elias_Saca 1 Peter_Holmberg 1 +Antonio_Elias_Saca 1 Thomas_Fargo 4 +Antonio_Trillanes 1 Katrin_Cartlidge 1 +Antonio_Trillanes 3 Terry_Gilliam 1 +Asa_Hutchinson 2 Gary_Sinise 1 +Ascencion_Barajas 1 Chris_Hernandez 1 +Ashley_Judd 1 Joan_Jett 1 +Ashley_Postell 1 Cassandra_Heise 1 +Ashley_Postell 1 Tony_Clement 1 +Ashraf_Ghani 1 Boris_Berezovsky 1 +Ashraf_Ghani 1 Qian_Qichen 1 +Asif_Ali_Zardari 1 Steny_Hoyer 1 +Assad_Ahmadi 1 Percy_Gibson 1 +Barbara_Walters 1 Jewel_Howard-Taylor 1 +Barbara_Walters 1 Peter_Greenaway 1 +Barry_Williams 1 Wang_Fei 1 +Benazir_Bhutto 4 Jane_Pauley 2 +Benazir_Bhutto 5 Karen_Clarkson 1 +Bijan_Darvish 1 Franco_Dragone 1 +Bijan_Darvish 2 Frank_Taylor 1 +Bilal_Erdogan 1 Lubomir_Zaoralek 1 +Bill_Self 1 Yang_Jianli 1 +Bill_Simon 1 Edward_Said 2 +Billy_Gilman 1 Howard_Smith 2 +Billy_Sollie 1 Mark_Swartz 1 +Billy_Sollie 2 Hussein_Malik 1 +Blas_Ople 1 Evan_Marriott 1 +Blas_Ople 1 Felipe_Fernandez 1 +Blas_Ople 1 Peter_Costello 1 +Bob_Cantrell 1 Gerard_Kleisterlee 1 +Bob_Cantrell 1 Mary_Landrieu 3 +Bob_Goldman 1 Leigh_Winchell 1 +Bob_Goldman 1 Maggie_Smith 1 +Boris_Berezovsky 2 Leander_Paes 1 +Brad_Gushue 1 Johannes_Rau 1 +Brad_Miller 1 Irina_Lobacheva 1 +Brad_Miller 1 Percy_Gibson 1 +Brenda_van_Dam 1 Norbert_van_Heyst 1 +Brian_Cashman 1 Delphine_Chuillot 1 +Brian_Cashman 1 Russell_Simmons 3 +Brian_Cook 1 Matt_Welsh 1 +Bruno_Junquiera 1 Oscar_DLeon 1 +Bulent_Ecevit 6 Eduardo_Romero 1 +Bulent_Ecevit 6 Gilberto_Simoni 1 +Candie_Kung 1 Steve_Allan 1 +Carlo_Ancelotti 1 John_Tyson 1 +Carlos_Vives 2 Randy_Johnson 1 +Caroline_Dhavernas 1 Evgeni_Plushenko 1 +Caroline_Dhavernas 1 Jeremy_Fogel 1 +Carson_Daly 1 John_Moe 1 +Carson_Daly 1 Tracy_Wyle 1 +Cassandra_Heise 1 Shaukat_Aziz 2 +Cate_Blanchett 3 Gina_Centrello 1 +Cate_Blanchett 3 Ryan_Goodman 1 +Catherine_Donkers 1 Eminem 1 +Cedric_Benson 1 Marcus_Allen 1 +Cedric_Benson 1 Robert_Lee_Yates_Jr 1 +Cedric_Benson 1 Stacy_Nelson 1 +Chistian_Stahl 1 Eliane_Karp 1 +Chris_Cooper 2 Irfan_Ahmed 1 +Chris_Cooper 2 Rick_Bland 1 +Chris_Hernandez 1 Mauricio_Pochetino 1 +Chris_Hernandez 1 Tonga 1 +Chris_Hernandez 1 Wolfgang_Schuessel 1 +Chris_Tucker 2 Jennifer_Pena 1 +Chris_Tucker 2 Takuma_Sato 1 +Chris_Whitney 1 Tom_Scully 1 +Christine_Gregoire 4 Robert_Lee_Yates_Jr 1 +Christine_Gregoire 4 Wang_Hailan 1 +Christine_Rau 1 Jerry_Hall 1 +Christine_Rau 1 Mel_Brooks 1 +Christopher_Russell 1 Colleen_OClair 1 +Christopher_Whittle 1 Felipe_Fernandez 1 +Colleen_OClair 1 Tom_Hanusik 1 +Curtis_Rodriguez 1 John_Blaney 2 +Curtis_Rodriguez 1 Kathleen_Kennedy_Townsend 3 +Dale_Bosworth 1 Frank_Taylor 1 +Dale_Bosworth 1 Vladimir_Golovlyov 1 +Dan_Quayle 1 Jim_Ryan 1 +Dan_Quayle 1 Nick_Turner 1 +Dan_Quayle 1 William_Umbach 1 +Danny_Morgan 1 Jeff_George 1 +Danny_Morgan 1 John_Allen_Muhammad 10 +Darla_Moore 1 Paul_Johnson 1 +Darrell_Royal 1 Louis_Van_Gaal 1 +Dave_McGinnis 1 Michel_Therrien 1 +David_Przybyszewski 1 Elena_Bovina 2 +Delphine_Chuillot 1 Jim_Parque 1 +Demetrin_Veal 1 Robert_Hanssen 1 +Dennis_Hastert 5 Lubomir_Zaoralek 1 +Dewayne_White 1 Joe_DeLamielleure 1 +Dick_Posthumus 1 Jeff_George 1 +Dick_Posthumus 1 Jim_Wong 1 +Din_Samsudin 1 Pupi_Avati 2 +Dionne_Warwick 1 Janine_Pietsch 1 +Dominique_Perben 1 Osmond_Smith 1 +Doug_Duncan 1 Kevin_Spacey 2 +Doug_Duncan 2 John_Allen_Muhammad 2 +Eddy_Hartenstein 1 Lyudmila_Putin 1 +Eddy_Hartenstein 1 Matthias_Sammer 1 +Eddy_Hartenstein 1 Sabah_Al-Ahmad_Al-Jaber_Al-Sabah 1 +Eduardo_Chillida 1 Stephen_Crampton 1 +Eduardo_Romero 1 Robert_Kocharian 5 +Edward_Greenspan 1 Roy_Halladay 1 +Edward_Greenspan 1 Sidney_Poitier 1 +Edward_Kennedy 1 Mathilda_Karel_Spak 1 +Edward_Kennedy 2 Jennifer_Pena 1 +Edward_Said 1 Frank_Abagnale_Jr 1 +Eminem 1 Wolfgang_Clement 1 +Enrique_Bolanos 4 Janine_Pietsch 1 +Enrique_Bolanos 4 Liam_Neeson 2 +Ernie_Els 2 Marcus_Allen 1 +Evgeni_Plushenko 1 Queen_Sofia 1 +Evgeni_Plushenko 1 Riek_Blanjaar 1 +Franco_Dragone 2 Louis_Van_Gaal 1 +Franco_Frattini 1 Irina_Lobacheva 1 +Frank_Abagnale_Jr 1 Nate_Hybl 1 +Frank_Shea 1 Rob_Moore 1 +Frank_Taylor 1 Jason_Jennings 1 +Gabriel_Batistuta 1 Peter_Costello 1 +Gabriel_Jorge_Ferreia 1 Zulfiqar_Ahmed 1 +Gary_Doer 3 Wang_Hailan 1 +Gary_Sinise 1 Roy_Halladay 1 +Gene_Hackman 1 Queen_Noor 1 +Gene_Sauers 1 Gina_Centrello 1 +Geoff_Dixon 1 Jean_Todt 1 +George_Allen 1 Roy_Halladay 1 +George_Allen 1 Wan_Yanhai 1 +George_Harrison 1 Robert_Lee_Yates_Jr 1 +Georgi_Parvanov 1 Luis_Fonsi 1 +Gianni_Agnelli 1 Marco_Irizarry 1 +Gilberto_Simoni 1 Julio_Cesar_Chavez 1 +Gong_Ruina 1 Hong_Myung 1 +Gong_Ruina 1 Qian_Qichen 1 +Gro_Harlem_Brundtland 2 Turner_Gill 1 +Hector_Grullon 1 Jeong_Se-hyun 9 +Henk_Bekedam 1 Koichi_Haraguchi 1 +Henk_Bekedam 1 Manuel_Gehring 1 +Henk_Bekedam 1 Michael_Kirby 1 +Hiroki_Gomi 1 Kenneth_Cooper 1 +Hiroki_Gomi 1 Sophia_Loren 2 +Irina_Lobacheva 1 Steve_Allan 1 +Ismail_Cem 1 Linda_Amicangioli 1 +Jack_Osbourne 1 Kent_McCord 1 +Jack_Straw 1 Manuel_Gehring 1 +Jack_Straw 3 Norman_Mineta 1 +James_Ballenger 1 Raza_Rabbani 1 +James_Coviello 1 Judith_Nathan 1 +James_Coviello 1 Robert_F_Kennedy_Jr 1 +James_Morris 2 Silvio_Fernandez 2 +James_Schultz 1 Tim_Curley 1 +James_W_Kennedy 1 John_Connolly 1 +James_W_Kennedy 1 Toutai_Kefu 1 +James_Williams 1 Phillip_Seymor_Hoffmann 1 +James_Williams 1 Shaun_Pollock 1 +Jan_Ullrich 3 Javier_Weber 1 +Jane_Menelaus 1 Junichiro_Koizumi 42 +Jane_Menelaus 1 Sophia_Loren 1 +Jason_Petty 1 Mayumi_Moriyama 1 +Jean_Todt 1 Roger_Etchegaray 1 +Jennifer_Pena 1 Sonia_Lopez 1 +Jennifer_Pena 1 Tony_LaRussa 1 +Jennifer_Rodriguez 1 Terry_Gilliam 1 +Jennifer_Rodriguez 2 Joanna_Poitier 1 +Jennifer_Rodriguez 2 Tony_Fernandes 1 +Jerome_Golmard 1 Kristin_Chenoweth 1 +Jewel_Howard-Taylor 1 Ken_Kutaragi 1 +Jim_Parque 1 Michael_Adams 1 +Jim_Parque 1 Rachel_Leigh_Cook 1 +Jim_Ryan 1 Luis_Berrondo 1 +Jimmy_Gobble 1 Robert_De_Niro 5 +Joan_Collins 1 Zydrunas_Ilgauskas 1 +Joan_Jett 1 Kevin_James 1 +Joanna_Poitier 1 Michael_Moore 1 +Joe_Lieberman 10 Joe_Mantello 1 +Johannes_Rau 1 Oleksandr_Moroz 1 +John_Connolly 1 Mel_Gibson 1 +John_Connolly 1 Michael_Moore 3 +John_Connolly 1 Norman_Mineta 1 +John_Connolly 1 Uzi_Even 1 +John_Cruz 1 Sean_Astin 3 +John_Cruz 1 Steven_Briggs 1 +John_Hartson 1 Oscar_Elias_Biscet 2 +Joseph_Deiss 1 Tony_Fernandes 1 +Joseph_Lopez 1 Rick_Pitino 1 +Julian_Fantino 1 Quin_Snyder 1 +Julio_Cesar_Chavez 1 Tom_Brennan 1 +Junichi_Inamoto 1 Oleg_Romantsev 1 +Junichiro_Koizumi 20 Tony_Fernandes 1 +Karen_Clarkson 1 Kristen_Breitweiser 3 +Karol_Kucera 1 Qian_Qichen 1 +Karol_Kucera 1 Turner_Gill 1 +Katrin_Cartlidge 1 Tony_LaRussa 1 +Keith_Osik 1 Pupi_Avati 1 +Kenneth_Cooper 1 Norman_Mineta 1 +Kent_McCord 1 Natasha_Henstridge 1 +Kevin_James 1 Roy_Halladay 1 +Kevin_James 1 Suzanne_Haik_Terrell 1 +Kevin_James 1 Yang_Jianli 1 +Kevin_Spacey 2 Ray_Evernham 1 +Koji_Uehara 1 Rob_Moore 1 +Krishna_Bhadur_Mahara 1 Peter_Holmberg 1 +Krishna_Bhadur_Mahara 1 Silvio_Fernandez 2 +Kristen_Breitweiser 1 Tatiana_Shchegoleva 1 +Laila_Ali 3 Tony_Fernandes 1 +Larry_Johnson 1 Terry_Gilliam 1 +Larry_Johnson 2 Marc_Racicot 1 +Larry_Johnson 2 Tom_Hanusik 1 +Lars_Von_Trier 1 Zhang_Wenkang 2 +Leander_Paes 1 Ralph_Firman 1 +Leo_Ramirez 1 Sabah_Al-Ahmad_Al-Jaber_Al-Sabah 1 +Leonardo_Del_Vecchio 1 Maggie_Smith 1 +Leonardo_Del_Vecchio 1 Tatiana_Shchegoleva 1 +Liam_Neeson 1 Paul_Kagame 2 +Lokendra_Bahadur_Chand 1 Steffeny_Holtz 1 +Louis_Van_Gaal 1 Natasha_Henstridge 1 +Lubomir_Zaoralek 1 Luis_Berrondo 1 +Lubomir_Zaoralek 1 William_Umbach 1 +Luis_Berrondo 1 Rita_Moreno 2 +Luis_Fonsi 1 Oscar_DLeon 1 +Luis_Fonsi 1 Steven_Briggs 1 +Luis_Rosario_Huertas 1 Mary_Steenburgen 2 +Makiko_Tanaka 1 Peter_Greenaway 1 +Manijeh_Hekmat 1 Roger_Winter 1 +Manijeh_Hekmat 1 T_Boone_Pickens 1 +Marcelo_Salas 2 Tara_VanDerveer 1 +Marco_Irizarry 1 Yang_Jianli 1 +Mariangel_Ruiz_Torrealba 3 Pupi_Avati 1 +Marisa_Tomei 2 Peter_Costello 2 +Mark_Polansky 1 Miranda_Gaddis 1 +Mary_Landrieu 3 Percy_Gibson 1 +Mathilda_Karel_Spak 1 Mauricio_Macri 1 +Mathilda_Karel_Spak 1 Tamara_Brooks 2 +Matt_Welsh 1 Randy_Johnson 1 +Mel_Brooks 2 Shoshana_Johnson 1 +Michael_Kirby 1 Ranil_Wickremasinghe 2 +Michael_Sheehan 1 Robert_Hanssen 1 +Michel_Duclos 1 Quin_Snyder 1 +Michel_Duclos 1 Suh_Young-hoon 1 +Mike_Gable 1 Mohamed_Hammam 1 +Mikhail_Khodorkovsky 1 Rita_Moreno 2 +Mireya_Moscoso 4 San_Lan 1 +Mohamed_Hammam 1 Roger_Moore 4 +Nabil_Shaath 3 Silvie_Cabero 1 +Naomi_Hayashi 1 Prakash_Hinduja 1 +Natalya_Sazanovich 1 Pupi_Avati 4 +Nikki_Teasley 1 Ryan_Goodman 1 +Oleksandr_Moroz 1 Robert_De_Niro 1 +Oscar_DLeon 1 Steven_Curtis_Chapman 1 +Paul_Byrd 1 Wolfgang_Schwarz 1 +Paul_Wollnough 1 Philip_Cummings 1 +Paul_Wollnough 1 Yuvraj_Singh 1 +Penelope_Taylor 1 Wang_Fei 1 +Peter_Caruana 1 Philip_Cummings 1 +Phillip_Seymor_Hoffmann 1 Silvio_Fernandez 1 +Phillips_Idowu 1 Prince_Naruhito 2 +Princess_Masako 2 Thomas_Wilkens 1 +Quin_Snyder 1 Vagit_Alekperov 1 +Ranil_Wickremasinghe 3 Stacy_Nelson 1 +Reina_Hayes 1 Steffeny_Holtz 1 +Rick_Pitino 3 Tony_Clement 1 +Ricky_Martin 2 Roger_Etchegaray 1 +Robert_F_Kennedy_Jr 1 Ron_Dittemore 1 +Sidney_Poitier 1 Svend_Aage_Jensby 1 +Abdel_Nasser_Assidi 1 2 +Ai_Sugiyama 1 2 +Ai_Sugiyama 1 4 +Ai_Sugiyama 4 5 +Aldo_Paredes 1 2 +Alejandro_Avila 1 2 +Alejandro_Avila 1 3 +Alex_Sink 2 3 +Allen_Iverson 1 2 +Amram_Mitzna 1 2 +Andrew_Niccol 1 2 +Andy_Hebb 1 2 +Anne_Krueger 1 2 +Anne_McLellan 1 3 +Anne_McLellan 2 3 +Annette_Bening 1 2 +Anthony_Hopkins 1 2 +Ariel_Sharon 16 45 +Arminio_Fraga 1 6 +Arminio_Fraga 2 4 +Arminio_Fraga 3 6 +Arminio_Fraga 4 6 +Art_Hoffmann 1 2 +Ashanti 1 3 +Ashanti 1 4 +Ashanti 2 5 +Ashanti 3 5 +Ashanti 4 5 +Augustin_Calleri 1 2 +Augustin_Calleri 2 4 +Augustin_Calleri 3 4 +Bertie_Ahern 1 2 +Bertie_Ahern 1 4 +Bertie_Ahern 1 5 +Bertie_Ahern 2 3 +Bertie_Ahern 3 5 +Bill_Clinton 2 4 +Bill_Clinton 6 8 +Bill_Clinton 6 10 +Bill_Clinton 8 10 +Bill_Clinton 9 12 +Bill_Clinton 10 29 +Bill_Clinton 18 28 +Bill_McBride 4 7 +Bill_McBride 4 10 +Bill_Parcells 1 2 +Binyamin_Ben-Eliezer 3 5 +Binyamin_Ben-Eliezer 5 6 +Binyamin_Ben-Eliezer 6 7 +Brendan_Hansen 1 2 +Brian_Wells 1 2 +Carlos_Quintanilla_Schmidt 1 2 +Carolina_Moraes 1 2 +Cecilia_Bolocco 1 2 +Cecilia_Bolocco 1 3 +Cecilia_Bolocco 2 3 +Cesar_Gaviria 2 4 +Cesar_Gaviria 2 5 +Cesar_Gaviria 3 6 +Cesar_Gaviria 3 7 +Cesar_Gaviria 4 6 +Charlie_Zaa 1 2 +Chita_Rivera 1 2 +Christian_Longo 1 2 +Christian_Longo 1 3 +Christian_Wulff 1 2 +Colin_Jackson 1 2 +Darrell_Issa 1 2 +Darrell_Porter 1 2 +Dave_Campo 1 3 +David_Dodge 1 2 +David_Heyman 1 2 +David_Spade 1 2 +David_Wells 1 5 +David_Wells 1 6 +David_Wells 1 7 +David_Wells 2 6 +David_Wells 5 6 +Debbie_Reynolds 1 3 +Debbie_Reynolds 2 4 +Dexter_Jackson 1 2 +Diana_Taurasi 1 2 +Donald_Evans 1 2 +Duane_Lee_Chapman 1 2 +Ed_Smart 1 3 +Eileen_Coparropa 1 2 +Eileen_Coparropa 1 3 +Eileen_Coparropa 2 3 +Elizabeth_Dole 1 3 +Elizabeth_Dole 2 4 +Fred_Funk 1 2 +Gabriel_Valdes 1 2 +Gary_Winnick 1 2 +Geno_Auriemma 1 2 +George_Karl 1 2 +George_Robertson 2 3 +George_Robertson 3 19 +George_Robertson 7 19 +George_Robertson 11 12 +George_Ryan 1 4 +George_Ryan 2 3 +George_W_Bush 7 65 +George_W_Bush 16 482 +George_W_Bush 145 150 +George_W_Bush 145 238 +George_W_Bush 203 247 +Gloria_Trevi 1 2 +Gloria_Trevi 2 3 +Gloria_Trevi 2 4 +Gordon_Campbell 1 2 +Greg_Rusedski 1 2 +Greg_Rusedski 2 3 +Greg_Rusedski 2 4 +Greg_Rusedski 3 4 +Gustavo_Kuerten 1 2 +Gustavo_Kuerten 2 3 +Guy_Hemmings 1 2 +Halle_Berry 6 7 +Halle_Berry 6 9 +Halle_Berry 7 9 +Heather_Mills 2 3 +Heather_Mills 2 4 +Heidi_Fleiss 1 2 +Heidi_Fleiss 1 3 +Henrique_Meirelles 1 2 +Iva_Majoli 1 2 +Jake_Gyllenhaal 3 5 +Jake_Gyllenhaal 4 5 +James_Cunningham 1 2 +James_Cunningham 1 3 +James_Cunningham 2 3 +James_Parker 1 2 +James_Wolfensohn 2 5 +Jason_Lezak 1 2 +Jay_Rasulo 1 2 +Jean-Claude_Juncker 1 2 +Jennifer_Lopez 1 3 +Jennifer_Lopez 2 16 +Jennifer_Lopez 3 10 +Jennifer_Lopez 3 14 +Jesse_Jackson 1 3 +Jesse_Jackson 5 8 +Jesse_Jackson 7 8 +Jesse_James_Leija 1 2 +Jesse_Ventura 2 3 +Jia_Qinglin 1 2 +Jim_Tressel 1 2 +Jim_Tressel 1 4 +Jim_Tressel 2 4 +Joan_Laporta 1 6 +Joan_Laporta 6 7 +Joan_Laporta 6 8 +Joe_Dumars 1 2 +John_Ashcroft 3 5 +John_Ashcroft 13 34 +John_Ashcroft 42 43 +John_Bolton 6 7 +John_Bolton 11 16 +John_Swofford 1 2 +John_Swofford 1 3 +John_Timoney 1 2 +Jon_Gruden 1 2 +Jon_Gruden 2 5 +Jon_Gruden 3 6 +Jon_Gruden 3 7 +Julie_Taymor 1 2 +Kamal_Kharrazi 1 5 +Kamal_Kharrazi 2 3 +Kamal_Kharrazi 2 6 +Karin_Stoiber 1 2 +Kim_Yong-il 1 2 +Kim_Yong-il 1 3 +Kimi_Raikkonen 1 3 +Kimi_Raikkonen 2 3 +Kjell_Magne_Bondevik 1 3 +Kobe_Bryant 1 2 +Kurt_Warner 1 2 +Kurt_Warner 1 4 +Kurt_Warner 2 4 +Kurt_Warner 3 5 +Kwon_Yang-sook 1 3 +Kwon_Yang-sook 2 3 +Lee_Jun 1 2 +Leonid_Kuchma 1 2 +Leonid_Kuchma 1 3 +Leonid_Kuchma 2 3 +Leonid_Kuchma 2 5 +Leonid_Kuchma 4 5 +Lina_Krasnoroutskaya 1 2 +Lisa_Raymond 1 2 +Lord_Hutton 1 2 +Luiz_Felipe_Scolari 1 2 +Lynn_Redgrave 2 3 +Magdalena_Maleeva 1 2 +Magdalena_Maleeva 1 3 +Magdalena_Maleeva 2 3 +Marcelo_Ebrard 1 2 +Marcelo_Ebrard 1 3 +Marcelo_Rios 1 3 +Marcelo_Rios 1 4 +Marcelo_Rios 2 3 +Marcelo_Rios 3 4 +Marcelo_Rios 3 5 +Maria_Shriver 1 7 +Maria_Shriver 1 8 +Maria_Shriver 2 6 +Maria_Shriver 2 8 +Maria_Shriver 3 8 +Marilyn_Monroe 1 2 +Mario_Cipollini 1 2 +Mark_Hurlbert 1 5 +Mark_Hurlbert 2 3 +Mark_Hurlbert 2 4 +Martin_Scorsese 2 5 +Martin_Scorsese 3 4 +Marwan_Barghouthi 1 2 +Michael_Bloomberg 1 12 +Michael_Bloomberg 3 5 +Michael_Bloomberg 7 11 +Michael_Bloomberg 11 13 +Michael_Chang 1 3 +Michael_Chang 5 6 +Michael_Jackson 1 6 +Michael_Jackson 2 3 +Michael_Jackson 2 11 +Michael_Jackson 2 12 +Michael_Jackson 11 12 +Michael_Patrick_King 1 2 +Michelle_Rodriguez 1 2 +Mike_Babcock 1 2 +Mike_Martz 1 3 +Mike_Scioscia 1 2 +Mikulas_Dzurinda 1 2 +Miroljub 1 2 +Nasser_al-Kidwa 1 2 +Nick_Nolte 1 2 +Nick_Nolte 1 3 +Nick_Nolte 1 5 +Nick_Nolte 2 4 +Nick_Nolte 3 4 +OJ_Simpson 1 2 +Oprah_Winfrey 1 2 +Oprah_Winfrey 2 3 +Orlando_Bloom 2 3 +Ozzy_Osbourne 1 3 +Paul_Coppin 1 2 +Paul_Martin 1 6 +Paul_Martin 1 7 +Paul_McNulty 1 2 +Paul_ONeill 1 3 +Paul_ONeill 5 7 +Paul_Patton 1 2 +Paula_Zahn 1 2 +Pierce_Brosnan 1 5 +Pierce_Brosnan 3 7 +Pierce_Brosnan 6 9 +Pierce_Brosnan 9 14 +Pierre_Pettigrew 1 3 +Ralf_Schumacher 5 8 +Ricky_Barnes 1 2 +Rita_Wilson 1 4 +Rob_Schneider 1 2 +Robbie_Fowler 1 2 +Robert_Blake 1 3 +Robert_Blake 1 4 +Robert_Blake 5 6 +Rogerio_Romero 1 2 +Rupert_Grint 1 3 +Salman_Rushdie 1 3 +Salman_Rushdie 2 3 +Scott_Sullivan 1 2 +Scott_Wolf 1 2 +Sepp_Blatter 1 2 +Sepp_Blatter 1 3 +Sepp_Blatter 2 3 +Simon_Cowell 1 2 +Slobodan_Milosevic 1 3 +Slobodan_Milosevic 2 3 +Slobodan_Milosevic 3 4 +Stacy_Dragila 1 2 +Stephen_Ambrose 1 2 +Stephen_Daldry 1 2 +Stephen_Friedman 1 2 +Theo_Epstein 1 2 +Thomas_Wyman 1 2 +Tim_Floyd 1 2 +Tom_Watson 1 3 +Tom_Watson 2 3 +Tommy_Robredo 1 3 +Tommy_Robredo 2 3 +Tony_Parker 1 2 +Tony_Stewart 1 6 +Tony_Stewart 2 3 +Tony_Stewart 4 5 +Tony_Stewart 4 6 +Vladimir_Voltchkov 1 2 +Wang_Yi 1 2 +Zafarullah_Khan_Jamali 1 2 +Zhu_Rongji 1 3 +Zhu_Rongji 2 8 +Aaron_Tippin 1 Enos_Slaughter 1 +Aaron_Tippin 1 Juan_Carlos_Ortega 1 +Aaron_Tippin 1 Marlon_Devonish 1 +Adam_Ant 1 John_Perrota 1 +Adam_Ant 1 Noel_Forgeard 1 +Adam_Ant 1 Richard_Regenhard 1 +Adam_Mair 1 Daniel_Osorno 1 +Adoor_Gopalakarishnan 1 Nathalia_Gillot 1 +Alain_Ducasse 1 Paul_ONeill 2 +Alan_Greer 1 Alan_Trammell 1 +Alan_Greer 1 Bob_Hayes 1 +Alan_Trammell 1 Heidi_Fleiss 2 +Alan_Trammell 1 Julie_Taymor 2 +Aldo_Paredes 1 Suzanne_Mubarak 1 +Aldo_Paredes 2 Zafarullah_Khan_Jamali 1 +Alecos_Markides 1 Darryl_Stingley 1 +Alecos_Markides 1 Wolfgang_Schneiderhan 1 +Alejandro_Avila 3 Benjamin_Franklin 1 +Alejandro_Avila 3 Darrell_Porter 1 +Alek_Wek 1 Barbara_Bodine 1 +Alek_Wek 1 Nasser_al-Kidwa 1 +Alek_Wek 1 Ray_Bradbury 1 +Alina_Kabaeva 1 Donald_Trump 1 +Alina_Kabaeva 1 Jayson_Williams 2 +Alyse_Beaupre 1 Elmar_Brok 1 +Alyse_Beaupre 1 Helo_Pinheiro 1 +Alyse_Beaupre 1 Hunter_Bates 1 +Ambrose_Lee 1 Ben_Chandler 1 +Amy_Gale 1 Herman_Edwards 1 +Anastasia_Kelesidou 1 Brendan_Hansen 2 +Anastasia_Kelesidou 1 Thomas_Wyman 2 +Andrew_Niccol 1 Prince_Felipe 1 +Andrew_Sabey 1 Federico_Castelan_Sayre 1 +Andrew_Sabey 1 Greg_Rusedski 1 +Andy_Madikians 1 Jon_Gruden 1 +Andy_Madikians 1 Pedro_Alvarez 1 +Andy_Madikians 1 Pierce_Brosnan 1 +Andy_Perez 1 Elena_Dementieva 1 +Andy_Perez 1 Stephen_Joseph 1 +Angie_Martinez 1 Ruth_Pearce 1 +Ann_Godbehere 1 Tom_Watson 1 +Anne_McLellan 1 Tim_Howard 1 +Annette_Bening 1 Ross_Verba 1 +Anthony_Hopkins 2 Orlando_Bloom 1 +Antonio_Bernardo 1 Billy_Donovan 1 +Antonio_Bernardo 1 Dwain_Kyles 1 +Antonio_Bernardo 1 Elizabeth_Hill 1 +Antonio_Bernardo 1 Guy_Hemmings 1 +Antonio_Bernardo 1 Kareena_Kapoor 1 +Anzori_Kikalishvili 1 Carlos_Quintanilla_Schmidt 2 +Anzori_Kikalishvili 1 Earl_Fritts 1 +Anzori_Kikalishvili 1 Salman_Rushdie 1 +Aram_Adler 1 Cesar_Gaviria 3 +Aram_Adler 1 Deepa_Mehta 1 +Arie_Haan 1 Tony_Parker 2 +Ariel_Sharon 30 David_Ballantyne 1 +Art_Hoffmann 2 Juan_Carlos_Ortega 1 +Atiabet_Ijan_Amabel 1 John_Perrota 1 +Augustin_Calleri 3 Lee_Jun 2 +Barry_Nakell 1 Maria_Simon 1 +Basdeo_Panday 1 Filippo_Volandri 1 +Ben_Betts 1 Kimi_Raikkonen 3 +Ben_Braun 1 Cecilia_Chang 1 +Ben_Braun 1 Horace_Newcomb 1 +Ben_Chandler 1 Larry_Hagman 1 +Ben_Lee 1 Horace_Newcomb 1 +Ben_Stein 1 David_Canary 1 +Ben_Stein 1 Lionel_Richie 2 +Bernadette_Peters 1 Ed_Smart 1 +Bertie_Ahern 4 Jim_Leach 1 +Bill_OReilly 1 Jim_Wessling 1 +Billy_Boyd 1 Sid_Caesar 1 +Billy_Donovan 1 Brandon_Spann 1 +Billy_Donovan 1 Cabas 1 +Billy_Donovan 1 William_Nessen 1 +Binyamin_Ben-Eliezer 1 Jenny_Romero 1 +Bo_Ryan 2 Heidi_Fleiss 3 +Bob_Beauprez 2 Hedayat_Amin_Arsala 1 +Bob_Beauprez 2 John_Norquist 1 +Bob_Hayes 1 Zhu_Rongji 8 +Brady_Rodgers 1 Jenna_Elfman 1 +Brandon_Larson 1 Chita_Rivera 1 +Brendan_Hansen 2 Dorothy_Wilson 1 +Caio_Blat 1 Sanja_Papic 1 +Carlos_Quintanilla_Schmidt 1 Nova_Esther_Guthrie 1 +Carlos_Quintanilla_Schmidt 2 Chris_Thomas 1 +Carroll_Weimer 1 Chris_Pronger 1 +Carroll_Weimer 1 Debbie_Reynolds 2 +Cecilia_Bolocco 1 John_Perrota 1 +Cecilia_Bolocco 1 Kurt_Schottenheimer 1 +Celia_Cruz 1 Rob_Ramsay 1 +Charlie_Zaa 2 Theo_Epstein 2 +Chris_Pronger 1 Edgar_Savisaar 1 +Christian_Longo 2 David_Carradine 1 +Christian_Wulff 2 Kjell_Magne_Bondevik 3 +Cindy_Klassen 1 Val_Ackerman 1 +Cindy_Taylor 1 Fabian_Vargas 1 +Claudette_Robinson 1 Eric_Fehr 1 +Clifford_Robinson 1 Magdalena_Maleeva 2 +Clifford_Robinson 1 Shirley_Jones 1 +Clifford_Robinson 1 Tim_Howard 1 +Colin_Jackson 1 Marcelo_Ebrard 3 +Colin_Jackson 2 Hunter_Bates 1 +Columba_Bush 1 Larry_Anderson 1 +Columba_Bush 1 Neil_Goldman 1 +Craig_Fitzgibbon 1 Joe_Dumars 1 +Cristian_Barros 1 Steve_Largent 1 +Curtis_Strange 1 Kurt_Schottenheimer 1 +Curtis_Strange 1 Raul_Chacon 1 +Dan_Kellner 1 Freda_Black 1 +Dan_LaCoutre 1 Gustavo_Kuerten 2 +Dan_Monson 1 Jim_Flaherty 1 +Dan_Monson 1 Rafael_Vinoly 1 +Darrell_Porter 2 Robert_Towne 1 +Dave_Campo 2 Jenny_Romero 1 +David_Ballantyne 1 Diana_Taurasi 2 +David_Carradine 1 Kobe_Bryant 1 +David_Dodge 1 Francois_Botha 1 +David_Dodge 1 Larry_Anderson 1 +David_Dodge 2 Thomas_Wyman 2 +David_Heyman 1 Rod_Paige 1 +David_Oh 1 Desiree_McKenzie 1 +David_Oh 1 Joe_Garner 1 +David_Spade 2 Denise_Locke 1 +David_Suazo 1 Jane_Krakowski 1 +David_Wells 1 Mary_Hill 1 +David_Wells 1 Yannos_Papantoniou 1 +Dawna_LoPiccolo 1 Jean-Claude_Juncker 2 +Debbie_Reynolds 1 Nick_Nolte 1 +Deniz_Baykal 1 Kathy_Bates 1 +Deniz_Baykal 1 Robin_Wagner 1 +Desiree_McKenzie 1 John_Danforth 1 +Dick_Devine 1 Michael_Jackson 9 +Dinora_Rosales 1 Robbie_Fowler 1 +Dirk_Kempthorne 1 Janusz_Kaminski 1 +Donald_Evans 2 Fred_Funk 2 +Dorothy_Wilson 1 Jim_Thome 1 +Drew_Gooden 1 Kimi_Raikkonen 3 +Dwain_Kyles 1 Gregory_Peck 1 +Ed_Smart 3 Jason_Gardner 1 +Edgar_Savisaar 1 Kirk_Doerger 1 +Edward_Burns 1 Paul_ONeill 1 +Elmar_Brok 1 Jose_Jose 1 +Emily_Mortimer 1 Maria_Simon 1 +Enos_Slaughter 1 Qais_al-Kazali 1 +Eric_Fehr 1 Lee_Jun 2 +Eric_Fehr 1 Tamara_Mowry 1 +Faisal_Saleh_Hayat 1 Marwan_Barghouthi 1 +Federico_Castelan_Sayre 1 Jimmy_Jimenez 1 +Fiona_Milne 1 Hartmut_Mehdorn 1 +Fiona_Milne 1 Josh_Kronfeld 1 +Floyd_Keith 1 Halle_Berry 1 +Franklin_Brown 1 Michael_Jackson 10 +Franklin_Brown 1 Tamara_Mowry 1 +Franklin_Brown 1 Yolanda_King 1 +Freda_Black 1 Rod_Paige 1 +Fruit_Chan 1 Kjell_Magne_Bondevik 3 +Gary_Winnick 1 Michael_Patrick_King 2 +George_Robertson 5 Robbie_Mc_Ewen 1 +Gloria_Trevi 4 William_Ragland 1 +Grace_Dodd 1 Henrique_Meirelles 1 +Graeme_Lloyd 1 Humberto_Coelho 1 +Graeme_Lloyd 1 Laura_Pausini 1 +Greg_Rusedski 2 Horace_Newcomb 1 +Greg_Rusedski 3 Javier_Bardem 1 +Greg_Rusedski 3 Marcos_Milinkovic 1 +Gregory_Peck 1 Jacqueline_Obradors 1 +Guillermo_Ruiz_Polanco 1 Jon_Constance 1 +Gustavo_Kuerten 1 Rani_Mukherjee 1 +Halle_Berry 12 Janet_Chandler 1 +Hartmut_Mehdorn 1 Natanaela_Barnova 1 +Hartmut_Mehdorn 1 Sanja_Papic 1 +Hartmut_Mehdorn 1 Svetlana_Belousova 1 +Haydar_Aliyev 1 Nuon_Chea 1 +Heather_Willson 1 Lynn_Redgrave 3 +Hector_Mitelman 1 Jon_Gruden 2 +Helo_Pinheiro 1 Robert_Durst 1 +Herman_Edwards 1 Patrick_Eaves 1 +Hermogenes_Ebdane_Jr 1 Sanja_Papic 1 +Horace_Newcomb 1 Stephan_Eberharter 1 +Hugo_Colace 1 Kim_Yun-kyu 1 +Humberto_Coelho 1 Sue_Wicks 2 +Hunter_Bates 1 Pedro_Alvarez 1 +Hunter_Bates 1 William_Rosenberg 1 +Igor_Trunov 1 Peter_Sejna 1 +Ivan_Stambolic 1 Michael_Bloomberg 18 +Jack_LaLanne 1 Yves_Brodeur 1 +Jacqueline_Obradors 1 Julie_Taymor 1 +Jada_Pinkett_Smith 2 Jenny_Romero 1 +James_Cunningham 3 Stephen_Daldry 1 +James_Gibson 1 Roger_Corbett 1 +James_Parker 1 Saeed_Mortazavi 1 +Jane_Krakowski 1 Shingo_Suetsugu 1 +Janice_Goldfinger 1 Pyar_Jung_Thapa 1 +Janusz_Kaminski 1 Lisa_Stone 1 +Jason_Lezak 1 John_Bolton 4 +Jason_Lezak 1 Simon_Cowell 1 +Javier_Vazquez 1 Karin_Stoiber 2 +Jay_Rasulo 1 Nova_Esther_Guthrie 1 +Jean-Marc_Olive 1 John_Timoney 1 +Jean-Marc_Olive 1 Tino_Martinez 1 +Jeri_Ryan 1 Nova_Esther_Guthrie 1 +Jeri_Ryan 1 Peter_Goldmark 1 +Jeri_Ryan 1 Scott_Sullivan 2 +Jeri_Ryan 1 Skip_Prosser 1 +Jesse_James_Leija 2 Joaquin_Phoenix 1 +Jesse_James_Leija 2 Nick_Nolte 4 +Jesse_Ventura 2 Lela_Rochon 1 +Jessica_Biel 1 Jim_Fassel 1 +Jim_Flaherty 1 Mark_Andrew 1 +Jim_Tressel 3 Ralf_Schumacher 3 +Jimmy_Jimenez 1 Miles_Stewart 1 +Jimmy_Lee 1 Nova_Esther_Guthrie 1 +Joan_Laporta 3 Roy_Romanow 1 +Joaquin_Phoenix 1 Rainer_Geulen 1 +Joe_Dumars 2 Timothy_McVeigh 1 +John_Bolton 14 Kim_Yun-kyu 1 +John_Kerr 1 Li_Changchun 1 +John_Swofford 2 Sok_An 1 +Juan_Fernandez 1 Paul_Krueger 1 +Juan_Fernandez 1 Suzanne_Mubarak 1 +Justin_Wilson 1 Larry_Hagman 1 +Justin_Wilson 1 Ray_Bradbury 1 +Kara_Lynn_Joyce 1 Zach_Pillar 1 +Kim_Chinn 1 Robert_Flodquist 1 +Kim_Chinn 1 Ruth_Harlow 2 +Kim_Yong-il 2 Linda_Ham 1 +Kwon_Yang-sook 3 Tommy_Robredo 1 +Lane_Bryant 1 Yannos_Papantoniou 1 +Larry_Hagman 1 Teddy_Kollek 1 +Larry_Hagman 1 Val_Ackerman 1 +Laura_Gobai 1 Robin_Wagner 1 +Laura_Pausini 1 Thad_Matta 1 +Laura_Romero 1 Paula_Zahn 2 +Lee_Jun 1 Ruth_Pearce 1 +Lee_Jun 2 Stephen_Friedman 1 +Lela_Rochon 1 Michael_Haneke 1 +Lela_Rochon 1 Yannos_Papantoniou 1 +Lesley_Flood 1 Peter_Goldmark 1 +Li_Changchun 1 Pieter_Bouw 1 +Lina_Krasnoroutskaya 1 Troy_Polamalu 1 +Linda_Lingle 1 Victor_Hanescu 1 +Lisa_Stone 1 Rod_Paige 1 +Lisa_Stone 1 Scott_Sullivan 2 +Lou_Lang 1 Nova_Esther_Guthrie 1 +Luc_Montagnier 1 Paul_Krueger 1 +Luc_Montagnier 1 Pedro_Alvarez 1 +Luc_Montagnier 1 Pierre_Pettigrew 2 +Luis_Guzman 1 Patsy_Kensit 1 +Marcelo_Ebrard 2 Mike_Scioscia 2 +Marcelo_Rios 4 Maria_Shriver 3 +Marcelo_Rios 4 Pierre_Pettigrew 1 +Marcelo_Rios 4 Saeed_Mortazavi 1 +Margaret_Thatcher 1 Pedro_Alvarez 1 +Margaret_Thatcher 1 Salman_Rushdie 2 +Maria_Simon 1 Mona_Rishmawi 1 +Maria_Simon 1 Robert_Durst 1 +Marlon_Devonish 1 Patrick_Clawsen 1 +Max_von_Sydow 1 Zhu_Rongji 1 +Mel_Karmazin 1 Pierre_Pettigrew 1 +Melana_Scantlin 1 William_Ragland 1 +Michael_Bloomberg 15 Ozzy_Osbourne 2 +Michael_J_Fox 1 Ricky_Barnes 1 +Michael_J_Fox 1 Thomas_Daily 1 +Micky_Ward 1 Takenori_Kanzaki 1 +Mike_Babcock 1 Tony_Parker 2 +Mike_Martz 5 William_Burns 1 +Mira_Sorvino 1 Simon_Cowell 2 +Mira_Sorvino 1 Tom_Tunney 1 +Miroljub 2 Theo_Epstein 1 +Mitzi_Gaynor 1 Ruth_Harlow 1 +Mohammed_Dahlan 1 Ronald_White 1 +Natanaela_Barnova 1 Nuon_Chea 1 +Noel_Forgeard 1 Zach_Pillar 1 +Normand_Legault 1 Omar_Vizquel 1 +Normand_Legault 1 Rupert_Grint 3 +Nova_Esther_Guthrie 1 Stephen_Joseph 1 +Ontario_Lett 1 Wallace_Capel 1 +Orlando_Bloom 1 Ray_Liotta 1 +Patrick_Clawsen 1 Sandra_Banning 1 +Paul_Coppin 2 Rick_Husband 1 +Paul_Murphy 1 Qazi_Hussain_Ahmed 1 +Paul_Newman 1 Robert_Blake 3 +Paula_Zahn 1 Tamara_Mowry 1 +Peter_Ahearn 1 Romain_Duris 1 +Peter_Gabriel 1 Peter_OToole 1 +Peter_Lundgren 1 William_Rosenberg 1 +Peter_OToole 1 Qazi_Afzal 1 +Qais_al-Kazali 1 Ringo_Starr 1 +Randy_Brown 1 Val_Ackerman 1 +Rani_Mukherjee 1 Timothy_McVeigh 1 +Ringo_Starr 1 Zach_Pillar 1 +Roger_Corbett 1 Tocker_Pudwill 1 +Ruth_Harlow 1 Virgina_Ruano_Pascal 1 +Sandra_Banning 1 Wolfgang_Schneiderhan 1 +Scott_Wolf 2 Troy_Polamalu 1 +Sergei_Alexandrovitch_Ordzhonikidze 1 Yolanda_King 1 +Shane_Loux 1 Val_Ackerman 1 +Shawn_Marion 1 Shirley_Jones 1 +Slobodan_Milosevic 2 Sok_An 1 diff --git a/TensorFlow/contrib/cv/CircleLoss_ID1221_for_TensorFlow/demo.py b/TensorFlow/contrib/cv/CircleLoss_ID1221_for_TensorFlow/demo.py deleted file mode 100644 index 714aeba9dc74b07e92f6162d3f0eed1cb0c3f4c8..0000000000000000000000000000000000000000 --- a/TensorFlow/contrib/cv/CircleLoss_ID1221_for_TensorFlow/demo.py +++ /dev/null @@ -1,138 +0,0 @@ -# Copyright 2017 The TensorFlow Authors. All Rights Reserved. -# -# Licensed under the Apache License, Version 2.0 (the "License"); -# you may not use this file except in compliance with the License. -# You may obtain a copy of the License at -# -# http://www.apache.org/licenses/LICENSE-2.0 -# -# Unless required by applicable law or agreed to in writing, software -# distributed under the License is distributed on an "AS IS" BASIS, -# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. -# See the License for the specific language governing permissions and -# limitations under the License. -# ============================================================================ -# Copyright 2021 Huawei Technologies Co., Ltd -# -# Licensed under the Apache License, Version 2.0 (the "License"); -# you may not use this file except in compliance with the License. -# You may obtain a copy of the License at -# -# http://www.apache.org/licenses/LICENSE-2.0 -# -# Unless required by applicable law or agreed to in writing, software -# distributed under the License is distributed on an "AS IS" BASIS, -# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. -# See the License for the specific language governing permissions and -# limitations under the License. - -import argparse -import scipy.io -import torch -import numpy as np -import os -from torchvision import datasets -import matplotlib -matplotlib.use('agg') -import matplotlib.pyplot as plt -####################################################################### -# Evaluate -parser = argparse.ArgumentParser(description='Demo') -parser.add_argument('--query_index', default=777, type=int, help='test_image_index') -parser.add_argument('--test_dir',default='/home/liulizhao/projects/liuyixin/Market-1501-v15.09.15/pytorch',type=str, help='./test_data') -opts = parser.parse_args() - -data_dir = opts.test_dir -image_datasets = {x: datasets.ImageFolder( os.path.join(data_dir,x) ) for x in ['gallery','query']} - -##################################################################### -#Show result -def imshow(path, title=None): - """Imshow for Tensor.""" - im = plt.imread(path) - plt.imshow(im) - if title is not None: - plt.title(title) - plt.pause(0.001) # pause a bit so that plots are updated - -###################################################################### -result = scipy.io.loadmat('/home/liulizhao/projects/liuyixin/projects/checkpoint/pytorch_result.mat') -query_feature = torch.FloatTensor(result['query_f']) -query_cam = result['query_cam'][0] -query_label = result['query_label'][0] -gallery_feature = torch.FloatTensor(result['gallery_f']) -gallery_cam = result['gallery_cam'][0] -gallery_label = result['gallery_label'][0] - -multi = os.path.isfile('/home/liulizhao/projects/liuyixin/projects/checkpoint/multi_query.mat') - -if multi: - m_result = scipy.io.loadmat('multi_query.mat') - mquery_feature = torch.FloatTensor(m_result['mquery_f']) - mquery_cam = m_result['mquery_cam'][0] - mquery_label = m_result['mquery_label'][0] - mquery_feature = mquery_feature.cuda() - -# query_feature = query_feature.cuda() -# gallery_feature = gallery_feature.cuda() - -####################################################################### -# sort the images -def sort_img(qf, ql, qc, gf, gl, gc): - query = qf.view(-1,1) - # print(query.shape) - score = torch.mm(gf,query) - score = score.squeeze(1).cpu() - score = score.numpy() - # predict index - index = np.argsort(score) #from small to large - index = index[::-1] - # index = index[0:2000] - # good index - query_index = np.argwhere(gl==ql) - #same camera - camera_index = np.argwhere(gc==qc) - - #good_index = np.setdiff1d(query_index, camera_index, assume_unique=True) - junk_index1 = np.argwhere(gl==-1) - junk_index2 = np.intersect1d(query_index, camera_index) - junk_index = np.append(junk_index2, junk_index1) - - mask = np.in1d(index, junk_index, invert=True) - index = index[mask] - return index - -i = opts.query_index -index = sort_img(query_feature[i],query_label[i],query_cam[i],gallery_feature,gallery_label,gallery_cam) - -######################################################################## -# Visualize the rank result - -query_path, _ = image_datasets['query'].imgs[i] -query_label = query_label[i] -print(query_path) -print('Top 10 images are as follow:') -try: # Visualize Ranking Result - # Graphical User Interface is needed - fig = plt.figure(figsize=(16,4)) - ax = plt.subplot(1,11,1) - ax.axis('off') - imshow(query_path,'query') - for i in range(10): - ax = plt.subplot(1,11,i+2) - ax.axis('off') - img_path, _ = image_datasets['gallery'].imgs[index[i]] - label = gallery_label[index[i]] - imshow(img_path) - if label == query_label: - ax.set_title('%d'%(i+1), color='green') - else: - ax.set_title('%d'%(i+1), color='red') - print(img_path) -except RuntimeError: - for i in range(10): - img_path = image_datasets.imgs[index[i]] - print(img_path[0]) - print('If you want to see the visualization of the ranking result, graphical user interface is needed.') - -fig.savefig("/home/liulizhao/projects/liuyixin/projects/show.png") diff --git a/TensorFlow/contrib/cv/CircleLoss_ID1221_for_TensorFlow/eval.py b/TensorFlow/contrib/cv/CircleLoss_ID1221_for_TensorFlow/eval.py deleted file mode 100644 index c2c69b14d1d09f61d2d79d85bb7d5eae263b363e..0000000000000000000000000000000000000000 --- a/TensorFlow/contrib/cv/CircleLoss_ID1221_for_TensorFlow/eval.py +++ /dev/null @@ -1,265 +0,0 @@ -# Copyright 2017 The TensorFlow Authors. All Rights Reserved. -# -# Licensed under the Apache License, Version 2.0 (the "License"); -# you may not use this file except in compliance with the License. -# You may obtain a copy of the License at -# -# http://www.apache.org/licenses/LICENSE-2.0 -# -# Unless required by applicable law or agreed to in writing, software -# distributed under the License is distributed on an "AS IS" BASIS, -# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. -# See the License for the specific language governing permissions and -# limitations under the License. -# ============================================================================ -# Copyright 2021 Huawei Technologies Co., Ltd -# -# Licensed under the Apache License, Version 2.0 (the "License"); -# you may not use this file except in compliance with the License. -# You may obtain a copy of the License at -# -# http://www.apache.org/licenses/LICENSE-2.0 -# -# Unless required by applicable law or agreed to in writing, software -# distributed under the License is distributed on an "AS IS" BASIS, -# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. -# See the License for the specific language governing permissions and -# limitations under the License. - -# #!/usr/bin/python -# # -*- encoding: utf-8 -*- -# import sys -# import os.path as osp -# import logging -# import pickle -# from tqdm import tqdm -# import numpy as np -# from data import loadMarketForTest - -# FORMAT = '%(levelname)s %(filename)s(%(lineno)d): %(message)s' -# logging.basicConfig(level=logging.INFO, format=FORMAT, stream=sys.stdout) -# logger = logging.getLogger(__name__) - -# # import os -# # os.environ['CUDA_VISBLE_DEVICES'] = '3' - -# import tensorflow as tf - -# sess=tf.Session() -# #先加载图和参数变量 -# saver = tf.train.import_meta_graph('/home/liulizhao/projects/liuyixin/projects/logs/fc2_adjustlr/model.ckpt-20500.meta') -# saver.restore(sess, tf.train.latest_checkpoint('/home/liulizhao/projects/liuyixin/projects/logs/fc2_adjustlr/')) -# # 访问placeholders变量，并且创建feed-dict来作为placeholders的新值 -# graph = tf.get_default_graph() -# x = graph.get_tensor_by_name("inputs:0") -# y = graph.get_tensor_by_name("labels:0") -# features = graph.get_tensor_by_name("features:0") - -# # 访问placeholders变量，并且创建feed-dict来作为placeholders的新值 -# graph = tf.get_default_graph() - -# def embed(): -# q,t = loadMarketForTest(data_dir='/home/liulizhao/projects/liuyixin/Market-1501-v15.09.15') -# query_embds, query_pids, query_camids = q.batch_sample(10,2,all=False) -# gallery_embds, gallery_pids, gallery_camids = t.batch_sample(750,3,all=False) - -# f1 = {x:query_embds,y:query_pids} -# query_embds = sess.run(features,f1) -# f2 = {x:gallery_embds,y:gallery_pids} -# gallery_embds = sess.run(features,f2) - -# embd_res = (query_embds, query_pids, query_camids,gallery_embds, gallery_pids, gallery_camids) - -# with open('/home/liulizhao/projects/liuyixin/projects/checkpoint/embds.pkl', 'wb') as fw: -# pickle.dump(embd_res, fw) -# # logger.info('embedding done, dump to: /home/liulizhao/projects/liuyixin/projects/checkpoint/embds.pkl') -# return embd_res - - -# def evaluate(embd_res, cmc_max_rank = 1): -# query_embds, query_pids, query_camids, gallery_embds, gallery_pids, gallery_camids = embd_res - -# ## compute distance matrix -# logger.info('compute distance matrix') -# dist_mtx = np.matmul(query_embds, gallery_embds.T) -# dist_mtx = 1.0 / (dist_mtx + 1) -# n_q, n_g = dist_mtx.shape - -# logger.info('start evaluating ...') -# indices = np.argsort(dist_mtx, axis = 1) -# matches = gallery_pids[indices] == query_pids[:, np.newaxis] -# matches = matches.astype(np.int32) -# all_aps = [] -# all_cmcs = [] -# for query_idx in tqdm(range(n_q)): -# query_pid = query_pids[query_idx] -# query_camid = query_camids[query_idx] - -# ## exclude duplicated gallery pictures -# order = indices[query_idx] -# pid_diff = gallery_pids[order] != query_pid -# camid_diff = gallery_camids[order] != query_camid -# useful = gallery_pids[order] != -1 -# keep = np.logical_or(pid_diff, camid_diff) -# keep = np.logical_and(keep, useful) -# match = matches[query_idx][keep] - -# if not np.any(match): continue - -# ## compute cmc -# cmc = match.cumsum() -# cmc[cmc > 1] = 1 -# all_cmcs.append(cmc[:cmc_max_rank]) - -# ## compute map -# num_real = match.sum() -# match_cum = match.cumsum() -# match_cum = [el / (1.0 + i) for i, el in enumerate(match_cum)] -# match_cum = np.array(match_cum) * match -# ap = match_cum.sum() / num_real -# all_aps.append(ap) - -# assert len(all_aps) > 0, "NO QUERRY APPEARS IN THE GALLERY" -# mAP = sum(all_aps) / len(all_aps) -# all_cmcs = np.array(all_cmcs, dtype = np.float32) -# cmc = np.mean(all_cmcs, axis = 0) - -# return cmc, mAP - - -# if __name__ == '__main__': -# embd_res = embed() -# with open('/home/liulizhao/projects/liuyixin/projects/checkpoint/embds.pkl', 'rb') as fr: -# embd_res = pickle.load(fr) - -# cmc, mAP = evaluate(embd_res) -# print('cmc is: {}, map is: {}'.format(cmc, mAP)) - - -import scipy.io -import torch -import numpy as np -#import time -import os - - -import argparse -parser = argparse.ArgumentParser(description='test') -parser.add_argument('--filename',default='pytorch_result.mat', type=str) -parser.add_argument('--logdir',default='pytorch_result.mat', type=str) -opt = parser.parse_args() - - -####################################################################### -# Evaluate -def evaluate(qf,ql,qc,gf,gl,gc): - query = qf.view(-1,1) - # print(query.shape) - score = torch.mm(gf,query) - score = score.squeeze(1).cpu() - score = score.numpy() - # predict index - index = np.argsort(score) #from small to large - index = index[::-1] - # index = index[0:2000] - # good index - query_index = np.argwhere(gl==ql) - camera_index = np.argwhere(gc==qc) - - good_index = np.setdiff1d(query_index, camera_index, assume_unique=True) - junk_index1 = np.argwhere(gl==-1) - junk_index2 = np.intersect1d(query_index, camera_index) - junk_index = np.append(junk_index2, junk_index1) #.flatten()) - - CMC_tmp = compute_mAP(index, good_index, junk_index) - return CMC_tmp - - -def compute_mAP(index, good_index, junk_index): - ap = 0 - cmc = torch.IntTensor(len(index)).zero_() - if good_index.size==0: # if empty - cmc[0] = -1 - return ap,cmc - - # remove junk_index - mask = np.in1d(index, junk_index, invert=True) - index = index[mask] - - # find good_index index - ngood = len(good_index) - mask = np.in1d(index, good_index) - rows_good = np.argwhere(mask==True) - rows_good = rows_good.flatten() - - cmc[rows_good[0]:] = 1 - for i in range(ngood): - d_recall = 1.0/ngood - precision = (i+1)*1.0/(rows_good[i]+1) - if rows_good[i]!=0: - old_precision = i*1.0/rows_good[i] - else: - old_precision=1.0 - ap = ap + d_recall*(old_precision + precision)/2 - - return ap, cmc - -###################################################################### -result = scipy.io.loadmat(opt.filename) -query_feature = torch.FloatTensor(result['query_f']) -query_cam = result['query_cam'][0] -query_label = result['query_label'][0] -gallery_feature = torch.FloatTensor(result['gallery_f']) -gallery_cam = result['gallery_cam'][0] -gallery_label = result['gallery_label'][0] - -multi = os.path.isfile('multi_query.mat') - -if multi: - m_result = scipy.io.loadmat('multi_query.mat') - mquery_feature = torch.FloatTensor(m_result['mquery_f']) - mquery_cam = m_result['mquery_cam'][0] - mquery_label = m_result['mquery_label'][0] - mquery_feature = mquery_feature.cuda() - -# query_feature = query_feature.cuda() -# gallery_feature = gallery_feature.cuda() - -print(query_feature.shape) -CMC = torch.IntTensor(len(gallery_label)).zero_() -ap = 0.0 -#print(query_label) -for i in range(len(query_label)): - ap_tmp, CMC_tmp = evaluate(query_feature[i],query_label[i],query_cam[i],gallery_feature,gallery_label,gallery_cam) - if CMC_tmp[0]==-1: - continue - CMC = CMC + CMC_tmp - ap += ap_tmp - #print(i, CMC_tmp[0]) - -CMC = CMC.float() -CMC = CMC/len(query_label) #average CMC -result_path = opt.logdir+"result.txt" -result_str = 'Rank@1:%f Rank@5:%f Rank@10:%f mAP:%f'%(CMC[0],CMC[4],CMC[9],ap/len(query_label)) -print(result_str) -with open(result_path,'a') as f: #设置文件对象 - f.write(result_str) #将字符串写入文件中 - -# multiple-query -CMC = torch.IntTensor(len(gallery_label)).zero_() -ap = 0.0 -if multi: - for i in range(len(query_label)): - mquery_index1 = np.argwhere(mquery_label==query_label[i]) - mquery_index2 = np.argwhere(mquery_cam==query_cam[i]) - mquery_index = np.intersect1d(mquery_index1, mquery_index2) - mq = torch.mean(mquery_feature[mquery_index,:], dim=0) - ap_tmp, CMC_tmp = evaluate(mq,query_label[i],query_cam[i],gallery_feature,gallery_label,gallery_cam) - if CMC_tmp[0]==-1: - continue - CMC = CMC + CMC_tmp - ap += ap_tmp - #print(i, CMC_tmp[0]) - CMC = CMC.float() - CMC = CMC/len(query_label) #average CMC - print('multi Rank@1:%f Rank@5:%f Rank@10:%f mAP:%f'%(CMC[0],CMC[4],CMC[9],ap/len(query_label))) diff --git a/TensorFlow/contrib/cv/CircleLoss_ID1221_for_TensorFlow/facenet.py b/TensorFlow/contrib/cv/CircleLoss_ID1221_for_TensorFlow/facenet.py new file mode 100644 index 0000000000000000000000000000000000000000..335eec2ee7cc4a3042edd03cd505b6372e51ca87 --- /dev/null +++ b/TensorFlow/contrib/cv/CircleLoss_ID1221_for_TensorFlow/facenet.py @@ -0,0 +1,598 @@ +# Copyright 2017 The TensorFlow Authors. All Rights Reserved. +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. +# ============================================================================ +# Copyright 2021 Huawei Technologies Co., Ltd +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. +from __future__ import absolute_import +from __future__ import division +from __future__ import print_function +from npu_bridge.npu_init import * + +import os +from subprocess import Popen, PIPE +import numpy as np +import tensorflow.compat.v1 as tf +tf.disable_eager_execution() +from scipy import misc +from sklearn.model_selection import KFold +from scipy import interpolate +import tensorflow.compat.v1.train as training +import random +import re +from tensorflow.compat.v1 import gfile +import math +from six import iteritems +import imageio +from tqdm import tqdm +from PIL import Image + + +def triplet_loss(anchor, positive, negative, alpha): + with tf.variable_scope('triplet_loss'): + pos_dist = tf.reduce_sum(tf.square(tf.subtract(anchor, positive)), 1) + neg_dist = tf.reduce_sum(tf.square(tf.subtract(anchor, negative)), 1) + + basic_loss = tf.add(tf.subtract(pos_dist, neg_dist), alpha) + loss = tf.reduce_mean(tf.maximum(basic_loss, 0.0), 0) + + return loss + + +def center_loss(features, label, alfa, nrof_classes): + + nrof_features = features.get_shape()[1] + centers = tf.get_variable('centers', [nrof_classes, nrof_features], dtype=tf.float32, + initializer=tf.constant_initializer(0), trainable=False) + label = tf.reshape(label, [-1]) + centers_batch = tf.gather(centers, label) + diff = (1 - alfa) * (centers_batch - features) + centers = tf.scatter_sub(centers, label, diff) + with tf.control_dependencies([centers]): + loss = tf.reduce_mean(tf.square(features - centers_batch)) + return loss, centers + + +def get_image_paths_and_labels(dataset): + image_paths_flat = [] + labels_flat = [] + for i in range(len(dataset)): + image_paths_flat += dataset[i].image_paths + labels_flat += [i] * len(dataset[i].image_paths) + return image_paths_flat, labels_flat + + +def shuffle_examples(image_paths, labels): + shuffle_list = list(zip(image_paths, labels)) + random.shuffle(shuffle_list) + image_paths_shuff, labels_shuff = zip(*shuffle_list) + return image_paths_shuff, labels_shuff + + +def random_rotate_image(image): + angle = np.random.uniform(low=-10.0, high=10.0) + return misc.imrotate(image, angle, 'bicubic') + +RANDOM_ROTATE = 1 +RANDOM_CROP = 2 +RANDOM_FLIP = 4 +FIXED_STANDARDIZATION = 8 +FLIP = 16 + +def create_input_pipeline(input_queue, image_size, nrof_preprocess_threads, batch_size_placeholder): + images_and_labels_list = [] + for _ in range(nrof_preprocess_threads): + filenames, label, control = input_queue.dequeue() + images = [] + for filename in tf.unstack(filenames): + file_contents = tf.read_file(filename) + image = tf.image.decode_image(file_contents, 3) + # image = tf.cast(image, tf.float32) + # image = tf.cond(get_control_flag(control[0], RANDOM_ROTATE), + # lambda: tf.py_func(random_rotate_image, [image], tf.uint8), + # lambda: tf.identity(image)) + image = tf.cond(get_control_flag(control[0], RANDOM_CROP), + lambda: tf.random_crop(image, image_size + (3,)), + lambda: tf.image.resize_image_with_crop_or_pad(image, image_size[0], image_size[1])) + image = tf.cond(get_control_flag(control[0], RANDOM_FLIP), + lambda: tf.image.random_flip_left_right(image), + lambda: tf.identity(image)) + image = tf.cond(get_control_flag(control[0], FIXED_STANDARDIZATION), + lambda: (tf.cast(image, tf.float32) - 127.5) / 128.0, + lambda: tf.cast(tf.image.per_image_standardization(image),tf.float32)) + image = tf.cond(get_control_flag(control[0], FLIP), + lambda: tf.image.flip_left_right(image), + lambda: tf.identity(image)) + # pylint: disable=no-member + image.set_shape(image_size + (3,)) + images.append(image) + images_and_labels_list.append([images, label]) + + image_batch, label_batch = tf.train.batch_join( + images_and_labels_list, batch_size=batch_size_placeholder, + shapes=[image_size + (3,), ()], enqueue_many=True, + capacity=4 * nrof_preprocess_threads * 100, + allow_smaller_final_batch=True) + + print("image_batch:",image_batch) + print("label_batch:",label_batch) + + return image_batch, label_batch + +def get_control_flag(control, field): + return tf.equal(tf.mod(tf.floor_div(control, field), 2), 1) + +def _add_loss_summaries(total_loss): + + loss_averages = tf.train.ExponentialMovingAverage(0.9, name='avg') + losses = tf.get_collection('losses') + loss_averages_op = loss_averages.apply(losses + [total_loss]) + + for l in losses + [total_loss]: + tf.summary.scalar(l.op.name + ' (raw)', l) + tf.summary.scalar(l.op.name, loss_averages.average(l)) + + return loss_averages_op + +def train(total_loss, global_step, optimizer, learning_rate, moving_average_decay, update_gradient_vars, + log_histograms=True): + + loss_averages_op = _add_loss_summaries(total_loss) + + with tf.control_dependencies([loss_averages_op]): + if optimizer == 'ADAGRAD': + opt = tf.train.AdagradOptimizer(learning_rate) + elif optimizer == 'ADADELTA': + opt = tf.train.AdadeltaOptimizer(learning_rate, rho=0.9, epsilon=1e-6) + elif optimizer == 'ADAM': + opt = tf.train.AdamOptimizer(learning_rate, beta1=0.9, beta2=0.999, epsilon=0.1) + elif optimizer == 'RMSPROP': + opt = tf.train.RMSPropOptimizer(learning_rate, decay=0.9, momentum=0.9, epsilon=1.0) + elif optimizer == 'MOM': + opt = tf.train.MomentumOptimizer(learning_rate, 0.9, use_nesterov=True) + elif optimizer == "GD": + opt = tf.train.GradientDescentOptimizer(learning_rate) + else: + raise ValueError('Invalid optimization algorithm') + + grads = opt.compute_gradients(total_loss, update_gradient_vars) + + apply_gradient_op = opt.apply_gradients(grads, global_step=global_step) + + if log_histograms: + for var in tf.trainable_variables(): + tf.summary.histogram(var.op.name, var) + + if log_histograms: + for grad, var in grads: + if grad is not None: + tf.summary.histogram(var.op.name + '/gradients', grad) + + variable_averages = tf.train.ExponentialMovingAverage( + moving_average_decay, global_step) + variables_averages_op = variable_averages.apply(tf.trainable_variables()) + + with tf.control_dependencies([apply_gradient_op, variables_averages_op]): + train_op = tf.no_op(name='train') + + return train_op + + +def prewhiten(x): + mean = np.mean(x) + std = np.std(x) + std_adj = np.maximum(std, 1.0 / np.sqrt(x.size)) + y = np.multiply(np.subtract(x, mean), 1 / std_adj) + return y + + +def crop(image, random_crop, image_size): + if image.shape[1] > image_size: + sz1 = int(image.shape[1] // 2) + sz2 = int(image_size // 2) + if random_crop: + diff = sz1 - sz2 + (h, v) = (np.random.randint(-diff, diff + 1), np.random.randint(-diff, diff + 1)) + else: + (h, v) = (0, 0) + image = image[(sz1 - sz2 + v):(sz1 + sz2 + v), (sz1 - sz2 + h):(sz1 + sz2 + h), :] + return image + + +def flip(image, random_flip): + if random_flip and np.random.choice([True, False]): + image = np.fliplr(image) + return image + + +def to_rgb(img): + w, h = img.shape + ret = np.empty((w, h, 3), dtype=np.uint8) + ret[:, :, 0] = ret[:, :, 1] = ret[:, :, 2] = img + return ret + +def load_data(image_paths, do_random_crop=False, do_random_flip=True, image_size=112, do_prewhiten=False): + nrof_samples = len(image_paths) + images = np.zeros((nrof_samples, image_size, image_size, 3)) + # (tf.cast(image, tf.float32) - 127.5) / 128.0, + for i in range(nrof_samples): + print("image_path:",image_paths[i]) + img = Image.open(image_paths[i]) + if img.ndim == 2: + img = to_rgb(img) + if do_prewhiten: + img = prewhiten(img) + if do_random_crop: + img = crop(img, do_random_crop, image_size) + if do_random_flip: + img = flip(img, do_random_flip) + + img=(img-127.5)/128.0 + images[i, :, :, :] = img + return images + + +def get_label_batch(label_data, batch_size, batch_index): + nrof_examples = np.size(label_data, 0) + j = batch_index * batch_size % nrof_examples + if j + batch_size <= nrof_examples: + batch = label_data[j:j + batch_size] + else: + x1 = label_data[j:nrof_examples] + x2 = label_data[0:nrof_examples - j] + batch = np.vstack([x1, x2]) + batch_int = batch.astype(np.int64) + return batch_int + + +def get_batch(image_data, batch_size, batch_index): + nrof_examples = np.size(image_data, 0) + j = batch_index * batch_size % nrof_examples + if j + batch_size <= nrof_examples: + batch = image_data[j:j + batch_size, :, :, :] + else: + x1 = image_data[j:nrof_examples, :, :, :] + x2 = image_data[0:nrof_examples - j, :, :, :] + batch = np.vstack([x1, x2]) + batch_float = batch.astype(np.float32) + return batch_float + + +def get_triplet_batch(triplets, batch_index, batch_size): + ax, px, nx = triplets + a = get_batch(ax, int(batch_size / 3), batch_index) + p = get_batch(px, int(batch_size / 3), batch_index) + n = get_batch(nx, int(batch_size / 3), batch_index) + batch = np.vstack([a, p, n]) + return batch + + +def get_learning_rate_from_file(filename, epoch): + with open(filename, 'r') as f: + for line in f.readlines(): + line = line.split('#', 1)[0] + if line: + par = line.strip().split(':') + e = int(par[0]) + if par[1] == '-': + lr = -1 + else: + lr = float(par[1]) + if e <= epoch: + learning_rate = lr + else: + return learning_rate + + +class ImageClass(): + def __init__(self, name, image_paths): + + self.name = name + + self.image_paths = image_paths + + def __str__(self): + + return self.name + ', ' + str(len(self.image_paths)) + ' images' + + def __len__(self): + + return len(self.image_paths) + + +def get_dataset(path, has_class_directories=True): + print("Get dataset.........") + dataset = [] + + path_exp = os.path.expanduser(path) + + classes = [path for path in tqdm(os.listdir(path_exp)) \ + if os.path.isdir(os.path.join(path_exp, path))] + + classes.sort() + + nrof_classes = len(classes) + + for i in tqdm(range(nrof_classes)): + + class_name = classes[i] + + facedir = os.path.join(path_exp, class_name) + + image_paths = get_image_paths(facedir) + + dataset.append(ImageClass(class_name, image_paths)) + return dataset + + +def get_image_paths(facedir): + image_paths = [] + if os.path.isdir(facedir): + images = os.listdir(facedir) + image_paths = [os.path.join(facedir, img) for img in images] + return image_paths + + +def split_dataset(dataset, split_ratio, min_nrof_images_per_class, mode): + if mode == 'SPLIT_CLASSES': + nrof_classes = len(dataset) + class_indices = np.arange(nrof_classes) + np.random.shuffle(class_indices) + split = int(round(nrof_classes * (1 - split_ratio))) + train_set = [dataset[i] for i in class_indices[0:split]] + test_set = [dataset[i] for i in class_indices[split:-1]] + elif mode == 'SPLIT_IMAGES': + train_set = [] + test_set = [] + for cls in dataset: + paths = cls.image_paths + np.random.shuffle(paths) + nrof_images_in_class = len(paths) + split = int(math.floor(nrof_images_in_class * (1 - split_ratio))) + if split == nrof_images_in_class: + split = nrof_images_in_class - 1 + if split >= min_nrof_images_per_class and nrof_images_in_class - split >= 1: + train_set.append(ImageClass(cls.name, paths[:split])) + test_set.append(ImageClass(cls.name, paths[split:])) + else: + raise ValueError('Invalid train/test split mode "%s"' % mode) + return train_set, test_set + + +def load_model(model, input_map=None): + + model_exp = os.path.expanduser(model) + if (os.path.isfile(model_exp)): + print('Model filename: %s' % model_exp) + with gfile.FastGFile(model_exp, 'rb') as f: + graph_def = tf.GraphDef() + graph_def.ParseFromString(f.read()) + tf.import_graph_def(graph_def, input_map=input_map, name='') + else: + print('Model directory: %s' % model_exp) + meta_file, ckpt_file = get_model_filenames(model_exp) + + print('Metagraph file: %s' % meta_file) + print('Checkpoint file: %s' % ckpt_file) + + saver = tf.train.import_meta_graph(os.path.join(model_exp, meta_file), input_map=input_map) + saver.restore(tf.get_default_session(), os.path.join(model_exp, ckpt_file)) + + +def get_model_filenames(model_dir): + files = os.listdir(model_dir) + meta_files = [s for s in files if s.endswith('.meta')] + if len(meta_files) == 0: + raise ValueError('No meta file found in the model directory (%s)' % model_dir) + elif len(meta_files) > 1: + raise ValueError('There should not be more than one meta file in the model directory (%s)' % model_dir) + meta_file = meta_files[0] + ckpt = tf.train.get_checkpoint_state(model_dir) + if ckpt and ckpt.model_checkpoint_path: + ckpt_file = os.path.basename(ckpt.model_checkpoint_path) + return meta_file, ckpt_file + + meta_files = [s for s in files if '.ckpt' in s] + max_step = -1 + for f in files: + step_str = re.match(r'(^model-[\w\- ]+.ckpt-(\d+))', f) + if step_str is not None and len(step_str.groups()) >= 2: + step = int(step_str.groups()[1]) + if step > max_step: + max_step = step + ckpt_file = step_str.groups()[0] + return meta_file, ckpt_file + + +def distance(embeddings1, embeddings2, distance_metric=0): + if distance_metric == 0: + + diff = np.subtract(embeddings1, embeddings2) + dist = np.sum(np.square(diff), 1) + elif distance_metric == 1: + + dot = np.sum(np.multiply(embeddings1, embeddings2), axis=1) + norm = np.linalg.norm(embeddings1, axis=1) * np.linalg.norm(embeddings2, axis=1) + similarity = dot / norm + dist = np.arccos(similarity) / math.pi + else: + raise 'Undefined distance metric %d' % distance_metric + + return dist + + +def calculate_roc(thresholds, embeddings1, embeddings2, actual_issame, nrof_folds=10, distance_metric=0, + subtract_mean=False): + assert (embeddings1.shape[0] == embeddings2.shape[0]) + assert (embeddings1.shape[1] == embeddings2.shape[1]) + nrof_pairs = min(len(actual_issame), embeddings1.shape[0]) + nrof_thresholds = len(thresholds) + k_fold = KFold(n_splits=nrof_folds, shuffle=False) + + tprs = np.zeros((nrof_folds, nrof_thresholds)) + fprs = np.zeros((nrof_folds, nrof_thresholds)) + accuracy = np.zeros((nrof_folds)) + + indices = np.arange(nrof_pairs) + + for fold_idx, (train_set, test_set) in enumerate(k_fold.split(indices)): + if subtract_mean: + mean = np.mean(np.concatenate([embeddings1[train_set], embeddings2[train_set]]), axis=0) + else: + mean = 0.0 + dist = distance(embeddings1 - mean, embeddings2 - mean, distance_metric) + + acc_train = np.zeros((nrof_thresholds)) + for threshold_idx, threshold in enumerate(thresholds): + _, _, acc_train[threshold_idx] = calculate_accuracy(threshold, dist[train_set], actual_issame[train_set]) + best_threshold_index = np.argmax(acc_train) + for threshold_idx, threshold in enumerate(thresholds): + tprs[fold_idx, threshold_idx], fprs[fold_idx, threshold_idx], _ = calculate_accuracy(threshold, + dist[test_set], + actual_issame[ + test_set]) + _, _, accuracy[fold_idx] = calculate_accuracy(thresholds[best_threshold_index], dist[test_set], + actual_issame[test_set]) + + tpr = np.mean(tprs, 0) + fpr = np.mean(fprs, 0) + return tpr, fpr, accuracy + + +def calculate_accuracy(threshold, dist, actual_issame): + predict_issame = np.less(dist, threshold) + tp = np.sum(np.logical_and(predict_issame, actual_issame)) + fp = np.sum(np.logical_and(predict_issame, np.logical_not(actual_issame))) + tn = np.sum(np.logical_and(np.logical_not(predict_issame), np.logical_not(actual_issame))) + fn = np.sum(np.logical_and(np.logical_not(predict_issame), actual_issame)) + + tpr = 0 if (tp + fn == 0) else float(tp) / float(tp + fn) + fpr = 0 if (fp + tn == 0) else float(fp) / float(fp + tn) + acc = float(tp + tn) / dist.size + return tpr, fpr, acc + + +def calculate_val(thresholds, embeddings1, embeddings2, actual_issame, far_target, nrof_folds=10, distance_metric=0, + subtract_mean=False): + assert (embeddings1.shape[0] == embeddings2.shape[0]) + assert (embeddings1.shape[1] == embeddings2.shape[1]) + nrof_pairs = min(len(actual_issame), embeddings1.shape[0]) + nrof_thresholds = len(thresholds) + k_fold = KFold(n_splits=nrof_folds, shuffle=False) + + val = np.zeros(nrof_folds) + far = np.zeros(nrof_folds) + + indices = np.arange(nrof_pairs) + + for fold_idx, (train_set, test_set) in enumerate(k_fold.split(indices)): + if subtract_mean: + mean = np.mean(np.concatenate([embeddings1[train_set], embeddings2[train_set]]), axis=0) + else: + mean = 0.0 + dist = distance(embeddings1 - mean, embeddings2 - mean, distance_metric) + + far_train = np.zeros(nrof_thresholds) + for threshold_idx, threshold in enumerate(thresholds): + _, far_train[threshold_idx] = calculate_val_far(threshold, dist[train_set], actual_issame[train_set]) + if np.max(far_train) >= far_target: + f = interpolate.interp1d(far_train, thresholds, kind='slinear') + threshold = f(far_target) + else: + threshold = 0.0 + + val[fold_idx], far[fold_idx] = calculate_val_far(threshold, dist[test_set], actual_issame[test_set]) + + val_mean = np.mean(val) + far_mean = np.mean(far) + val_std = np.std(val) + return val_mean, val_std, far_mean + + +def calculate_val_far(threshold, dist, actual_issame): + predict_issame = np.less(dist, threshold) + true_accept = np.sum(np.logical_and(predict_issame, actual_issame)) + false_accept = np.sum(np.logical_and(predict_issame, np.logical_not(actual_issame))) + n_same = np.sum(actual_issame) + n_diff = np.sum(np.logical_not(actual_issame)) + val = float(true_accept) / float(n_same) + far = float(false_accept) / float(n_diff) + return val, far + + +def store_revision_info(src_path, output_dir, arg_string): + try: + cmd = ['git', 'rev-parse', 'HEAD'] + gitproc = Popen(cmd, stdout=PIPE, cwd=src_path) + (stdout, _) = gitproc.communicate() + git_hash = stdout.strip() + except OSError as e: + git_hash = ' '.join(cmd) + ': ' + e.strerror + + try: + cmd = ['git', 'diff', 'HEAD'] + gitproc = Popen(cmd, stdout=PIPE, cwd=src_path) + (stdout, _) = gitproc.communicate() + git_diff = stdout.strip() + except OSError as e: + git_diff = ' '.join(cmd) + ': ' + e.strerror + + rev_info_filename = os.path.join(output_dir, 'revision_info.txt') + with open(rev_info_filename, "w") as text_file: + text_file.write('arguments: %s\n--------------------\n' % arg_string) + text_file.write('tensorflow version: %s\n--------------------\n' % tf.__version__) + text_file.write('git hash: %s\n--------------------\n' % git_hash) + text_file.write('%s' % git_diff) + + +def list_variables(filename): + reader = training.NewCheckpointReader(filename) + variable_map = reader.get_variable_to_shape_map() + names = sorted(variable_map.keys()) + return names + + +def put_images_on_grid(images, shape=(16, 8)): + nrof_images = images.shape[0] + img_size = images.shape[1] + bw = 3 + img = np.zeros((shape[1] * (img_size + bw) + bw, shape[0] * (img_size + bw) + bw, 3), np.float32) + for i in range(shape[1]): + x_start = i * (img_size + bw) + bw + for j in range(shape[0]): + img_index = i * shape[0] + j + if img_index >= nrof_images: + break + y_start = j * (img_size + bw) + bw + img[x_start:x_start + img_size, y_start:y_start + img_size, :] = images[img_index, :, :, :] + if img_index >= nrof_images: + break + return img + + +def write_arguments_to_file(args, filename): + with open(filename, 'w') as f: + for key, value in iteritems(vars(args)): + f.write('%s: %s\n' % (key, str(value))) + diff --git a/TensorFlow/contrib/cv/CircleLoss_ID1221_for_TensorFlow/config.py b/TensorFlow/contrib/cv/CircleLoss_ID1221_for_TensorFlow/lfw.py similarity index 33% rename from TensorFlow/contrib/cv/CircleLoss_ID1221_for_TensorFlow/config.py rename to TensorFlow/contrib/cv/CircleLoss_ID1221_for_TensorFlow/lfw.py index 8203193a3ce8c2d56734f3477bf0cfc49ccc2e13..20d57f00ff510a202a0c20969b7b0298d04c7449 100644 --- a/TensorFlow/contrib/cv/CircleLoss_ID1221_for_TensorFlow/config.py +++ b/TensorFlow/contrib/cv/CircleLoss_ID1221_for_TensorFlow/lfw.py @@ -25,46 +25,68 @@ # WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. # See the License for the specific language governing permissions and # limitations under the License. +from __future__ import absolute_import +from __future__ import division +from __future__ import print_function +from npu_bridge.npu_init import * -import argparse -arg = argparse.ArgumentParser() -import time -timestamp = int(time.time()) -arg.add_argument('-e','--expriment_name',type=str,default=f'test_{timestamp}') -arg.add_argument('-g','--gpu',type=str,default='0,1') -arg.add_argument('-w','--warmup',type=int,default=5) -arg.add_argument('-b','--batch_size',type=int,default=32) -arg.add_argument('-epoch','--epoch_size',type=int,default=120) -arg.add_argument('-modelname','--modelname',type=str,default='softmaxcircleloss') -arg.add_argument('-lr','--init_lr',type=float,default=0.02) -arg.add_argument('-lamda','--lamda',type=float,default=1) -arg.add_argument('-drop','--droprate',type=float,default=0.5) +import os +import numpy as np +import facenet -# for huawei npu -arg.add_argument("--train_url", type=str, default="./output") -arg.add_argument("--data_url", type=str, default="./dataset") -arg.add_argument("--ckpt_url", type=str, default="./ckpt") -args = arg.parse_args() -# args.expriment_name = args.expriment_name+"_"+str(timestamp) -homedir = "/home/nanshen/xutan/yixin/circlesoftmax" -log_dir = f'{homedir}/logs/{args.expriment_name}/' -im_size = [256, 128] -market_dir = 'Data/Market-1501-v15.09.15' -# P=4 -warmup = args.warmup -# K=4 -m=0.25 -gamma=32 -lamda = args.lamda -num_classes=751 -epoch_size = args.epoch_size -batch_size = args.batch_size -embedding_size=128 -resnet_model_path = '/home/nanshen/xutan/yixin/resnet_v1_50.ckpt' # Path to the pretrained model -model_save_path = log_dir + args.modelname # Path to the model.ckpt-(num_steps) will be saved -num_bottleneck = 512 -import os -os.environ["CUDA_VISIBLE_DEVICES"]=args.gpu -init_lr = args.init_lr +def evaluate(embeddings, actual_issame, nrof_folds=10, distance_metric=0, subtract_mean=False): + thresholds = np.arange(0, 4, 0.01) + embeddings1 = embeddings[0::2] + embeddings2 = embeddings[1::2] + tpr, fpr, accuracy = facenet.calculate_roc(thresholds, embeddings1, embeddings2, + np.asarray(actual_issame), nrof_folds=nrof_folds, + distance_metric=distance_metric, subtract_mean=subtract_mean) + thresholds = np.arange(0, 4, 0.001) + val, val_std, far = facenet.calculate_val(thresholds, embeddings1, embeddings2, + np.asarray(actual_issame), 1e-3, nrof_folds=nrof_folds, + distance_metric=distance_metric, subtract_mean=subtract_mean) + return tpr, fpr, accuracy, val, val_std, far + + +def get_paths(lfw_dir, pairs): + nrof_skipped_pairs = 0 + path_list = [] + issame_list = [] + for pair in pairs: + if len(pair) == 3: + path0 = add_extension(os.path.join(lfw_dir, pair[0], pair[0] + '_' + '%04d' % int(pair[1]))) + path1 = add_extension(os.path.join(lfw_dir, pair[0], pair[0] + '_' + '%04d' % int(pair[2]))) + issame = True + elif len(pair) == 4: + path0 = add_extension(os.path.join(lfw_dir, pair[0], pair[0] + '_' + '%04d' % int(pair[1]))) + path1 = add_extension(os.path.join(lfw_dir, pair[2], pair[2] + '_' + '%04d' % int(pair[3]))) + issame = False + if os.path.exists(path0) and os.path.exists(path1): # Only add the pair if both paths exist + path_list += (path0, path1) + issame_list.append(issame) + else: + nrof_skipped_pairs += 1 + if nrof_skipped_pairs > 0: + print('Skipped %d image pairs' % nrof_skipped_pairs) + + return path_list, issame_list + + +def add_extension(path): + if os.path.exists(path + '.jpg'): + return path + '.jpg' + elif os.path.exists(path + '.png'): + return path + '.png' + else: + raise RuntimeError('No file "%s" with extension png or jpg.' % path) + + +def read_pairs(pairs_filename): + pairs = [] + with open(pairs_filename, 'r') as f: + for line in f.readlines()[1:]: + pair = line.strip().split() + pairs.append(pair) + return np.array(pairs) diff --git a/TensorFlow/contrib/cv/CircleLoss_ID1221_for_TensorFlow/loss.png b/TensorFlow/contrib/cv/CircleLoss_ID1221_for_TensorFlow/loss.png new file mode 100644 index 0000000000000000000000000000000000000000..897155bdfe0357adf978f94e53801c971a8792fc Binary files /dev/null and b/TensorFlow/contrib/cv/CircleLoss_ID1221_for_TensorFlow/loss.png differ diff --git a/TensorFlow/contrib/cv/CircleLoss_ID1221_for_TensorFlow/loss.py b/TensorFlow/contrib/cv/CircleLoss_ID1221_for_TensorFlow/loss.py deleted file mode 100644 index 5256e55f77fc9a46d71287975cbee1e6cf12eb34..0000000000000000000000000000000000000000 --- a/TensorFlow/contrib/cv/CircleLoss_ID1221_for_TensorFlow/loss.py +++ /dev/null @@ -1,170 +0,0 @@ -# Copyright 2017 The TensorFlow Authors. All Rights Reserved. -# -# Licensed under the Apache License, Version 2.0 (the "License"); -# you may not use this file except in compliance with the License. -# You may obtain a copy of the License at -# -# http://www.apache.org/licenses/LICENSE-2.0 -# -# Unless required by applicable law or agreed to in writing, software -# distributed under the License is distributed on an "AS IS" BASIS, -# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. -# See the License for the specific language governing permissions and -# limitations under the License. -# ============================================================================ -# Copyright 2021 Huawei Technologies Co., Ltd -# -# Licensed under the Apache License, Version 2.0 (the "License"); -# you may not use this file except in compliance with the License. -# You may obtain a copy of the License at -# -# http://www.apache.org/licenses/LICENSE-2.0 -# -# Unless required by applicable law or agreed to in writing, software -# distributed under the License is distributed on an "AS IS" BASIS, -# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. -# See the License for the specific language governing permissions and -# limitations under the License. - -import tensorflow as tf -# from config import * -from train_npuv3 import * -from model import weight_variable -import numpy as np -def feature2similaryMatrix(y_norm): - return tf.matmul(y_norm,tf.transpose(y_norm)) - -def reducelogsumexp(x): - x = tf.math.exp(x) - x = tf.math.reduce_sum(x) - x = tf.math.log(x) - return x - -def PairWiseCircleLossv2(y_norm, y_label): - label_matrix = tf.equal(tf.expand_dims(y_label, 1), tf.expand_dims(y_label, 0)) - # label_matrix = tf.cast(label_matrix, tf.int64) - - similarity_matrix = tf.matmul(y_norm, tf.transpose(y_norm)) - # negative_matrix = tf.linalg.band_part(tf.logical_not(tf.cast(label_matrix, tf.bool)), 0, -1) - # positive_matrix_1 = label_matrix - tf.linalg.band_part(label_matrix, 0, 0) - # positive_matrix = tf.linalg.band_part(positive_matrix_1, 0, -1) - # positive_matrix = tf.cast(positive_matrix, tf.bool) - # positive_matrix = - - similarity_matrix_flat = tf.reshape(similarity_matrix, [-1]) - same = tf.reshape(label_matrix, [-1]) - dif = tf.logical_not(same) - sp = similarity_matrix_flat[same] - sn = similarity_matrix_flat[dif] - - # 计算代价函数 - op = 1 + m - on = -m - ap = tf.clip_by_value(op - sp, clip_value_min=0., clip_value_max=float('inf')) - an = tf.clip_by_value(sn - on, clip_value_min=0., clip_value_max=float('inf')) - delta_p = 1 - m - delta_n = m - logit_p = - ap * (sp - delta_p) * gamma - logit_n = an * (sn - delta_n) * gamma - loss = tf.math.softplus(tf.math.reduce_logsumexp(logit_n) + tf.math.reduce_logsumexp(logit_p)) - - recall, precision, f1_score = get_metric(m, similarity_matrix, label_matrix) - - return loss, sp, sn, recall, precision, f1_score - -def PairWiseCircleLoss(y_norm, y_label): - label_matrix = tf.equal(tf.expand_dims(y_label, 1), tf.expand_dims(y_label, 0)) - label_matrix = tf.cast(label_matrix, tf.int64) - - similarity_matrix = tf.matmul(y_norm, tf.transpose(y_norm)) - negative_matrix = tf.linalg.band_part(tf.logical_not(tf.cast(label_matrix, tf.bool)), 0, -1) - positive_matrix_1 = label_matrix - tf.linalg.band_part(label_matrix, 0, 0) - positive_matrix = tf.linalg.band_part(positive_matrix_1, 0, -1) - positive_matrix = tf.cast(positive_matrix, tf.bool) - - similarity_matrix_flat = tf.reshape(similarity_matrix, [-1]) - positive_matrix = tf.reshape(positive_matrix, [-1]) - negative_matrix = tf.reshape(negative_matrix, [-1]) - - sp = tf.cond(tf.equal(tf.size(positive_matrix), 0), lambda: 0.0, lambda: similarity_matrix_flat[positive_matrix]) - sn = tf.cond(tf.equal(tf.size(negative_matrix), 0), lambda: 0.0, lambda: similarity_matrix_flat[negative_matrix]) - - # 计算代价函数 - op = 1 + m - on = -m - ap = tf.clip_by_value(op - sp, clip_value_min=0., clip_value_max=float('inf')) - an = tf.clip_by_value(sn - on, clip_value_min=0., clip_value_max=float('inf')) - delta_p = 1 - m - delta_n = m - logit_p = - ap * (sp - delta_p) * gamma - logit_n = an * (sn - delta_n) * gamma - loss = tf.math.softplus(reducelogsumexp(logit_n) + reducelogsumexp(logit_p)) - - recall, precision, f1_score = get_metric(m, similarity_matrix, label_matrix) - - return loss, sp, sn, recall, precision, f1_score - - -# 取矩阵的上三角矩阵，对角线置False，转bool类型 -def get_uppertrimatrix(tensor): - tensor = tf.cast(tensor,tf.float32) - tensor = tensor - tf.linalg.band_part(tensor,0,0) - tensor = tf.linalg.band_part(tensor,0,-1) - tensor = tf.cast(tensor,tf.bool) - return tensor - -def get_metric(m,similarity_matrix,label_matrix): - thresh = 0.5 - K = get_uppertrimatrix(similarity_matrix>=thresh) - K_bar = get_uppertrimatrix(similarity_matrix 1). - """ - if factor == 1: - return inputs - else: - return slim.max_pool2d(inputs, [1, 1], stride=factor, scope=scope) - - -def conv2d_same(inputs, num_outputs, kernel_size, stride, scope=None): - """Strided 2-D convolution with 'SAME' padding. - When stride > 1, then we do explicit zero-padding, followed by conv2d with - 'VALID' padding. - Note that - net = conv2d_same(inputs, num_outputs, 3, stride=stride) - is equivalent to - net = slim.conv2d(inputs, num_outputs, 3, stride=1, padding='SAME') - net = subsample(net, factor=stride) - whereas - net = slim.conv2d(inputs, num_outputs, 3, stride=stride, padding='SAME') - is different when the input's height or width is even, which is why we add the - current function. For more details, see ResnetUtilsTest.testConv2DSameEven(). - Args: - inputs: A 4-D tensor of size [batch, height_in, width_in, channels]. - num_outputs: An integer, the number of output filters. - kernel_size: An int with the kernel_size of the filters. - stride: An integer, the output stride. - rate: An integer, rate for atrous convolution. - scope: Scope. - Returns: - output: A 4-D tensor of size [batch, height_out, width_out, channels] with - the convolution output. - """ - if stride == 1: - return slim.conv2d(inputs, num_outputs, kernel_size, stride=1, - padding='SAME', scope=scope) - else: - #kernel_size_effective = kernel_size + (kernel_size - 1) * (rate - 1) - pad_total = kernel_size - 1 - pad_beg = pad_total // 2 - pad_end = pad_total - pad_beg - inputs = tf.pad(inputs, - [[0, 0], [pad_beg, pad_end], [pad_beg, pad_end], [0, 0]]) - return slim.conv2d(inputs, num_outputs, kernel_size, stride=stride, - padding='VALID', scope=scope) - - -@slim.add_arg_scope -def stack_blocks_dense(net, blocks, - outputs_collections=None): - """Stacks ResNet `Blocks` and controls output feature density. - First, this function creates scopes for the ResNet in the form of - 'block_name/unit_1', 'block_name/unit_2', etc. - Args: - net: A `Tensor` of size [batch, height, width, channels]. - blocks: A list of length equal to the number of ResNet `Blocks`. Each - element is a ResNet `Block` object describing the units in the `Block`. - outputs_collections: Collection to add the ResNet block outputs. - Returns: - net: Output tensor - """ - for block in blocks: - with tf.variable_scope(block.scope, 'block', [net]) as sc: - for i, unit in enumerate(block.args): - - with tf.variable_scope('unit_%d' % (i + 1), values=[net]): - unit_depth, unit_depth_bottleneck, unit_stride = unit - net = block.unit_fn(net, - depth=unit_depth, - depth_bottleneck=unit_depth_bottleneck, - stride=unit_stride) - net = slim.utils.collect_named_outputs(outputs_collections, sc.name, net) - - return net - - -def resnet_arg_scope(is_training=True, - weight_decay=0.0001, - batch_norm_decay=0.997, - batch_norm_epsilon=1e-5, - batch_norm_scale=True): - """Defines the default ResNet arg scope. - TODO(gpapan): The batch-normalization related default values above are - appropriate for use in conjunction with the reference ResNet models - released at https://github.com/KaimingHe/deep-residual-networks. When - training ResNets from scratch, they might need to be tuned. - Args: - is_training: Whether or not we are training the parameters in the batch - normalization layers of the model. - weight_decay: The weight decay to use for regularizing the model. - batch_norm_decay: The moving average decay when estimating layer activation - statistics in batch normalization. - batch_norm_epsilon: Small constant to prevent division by zero when - normalizing activations by their variance in batch normalization. - batch_norm_scale: If True, uses an explicit `gamma` multiplier to scale the - activations in the batch normalization layer. - Returns: - An `arg_scope` to use for the resnet models. - """ - batch_norm_params = { - 'is_training': is_training, - 'decay': batch_norm_decay, - 'epsilon': batch_norm_epsilon, - 'scale': batch_norm_scale, - 'updates_collections': tf.GraphKeys.UPDATE_OPS, - } - - with slim.arg_scope( - [slim.conv2d], - weights_regularizer=slim.l2_regularizer(weight_decay), - weights_initializer=slim.variance_scaling_initializer(), - activation_fn=tf.nn.relu, - normalizer_fn=slim.batch_norm, - normalizer_params=batch_norm_params): - with slim.arg_scope([slim.batch_norm], **batch_norm_params): - # The following implies padding='SAME' for pool1, which makes feature - # alignment easier for dense prediction tasks. This is also used in - # https://github.com/facebook/fb.resnet.torch. However the accompanying - # code of 'Deep Residual Learning for Image Recognition' uses - # padding='VALID' for pool1. You can switch to that choice by setting - # slim.arg_scope([slim.max_pool2d], padding='VALID'). - with slim.arg_scope([slim.max_pool2d], padding='SAME') as arg_sc: - return arg_sc - - - - -@slim.add_arg_scope -def bottleneck(inputs, depth, depth_bottleneck, stride, - outputs_collections=None, scope=None): - """Bottleneck residual unit variant with BN before convolutions. - This is the full preactivation residual unit variant proposed in [2]. See - Fig. 1(b) of [2] for its definition. Note that we use here the bottleneck - variant which has an extra bottleneck layer. - When putting together two consecutive ResNet blocks that use this unit, one - should use stride = 2 in the last unit of the first block. - Args: - inputs: A tensor of size [batch, height, width, channels]. - depth: The depth of the ResNet unit output. - depth_bottleneck: The depth of the bottleneck layers. - stride: The ResNet unit's stride. Determines the amount of downsampling of - the units output compared to its input. - rate: An integer, rate for atrous convolution. - outputs_collections: Collection to add the ResNet unit output. - scope: Optional variable_scope. - Returns: - The ResNet unit's output. - """ - with tf.variable_scope(scope, 'bottleneck_v2', [inputs]) as sc: - depth_in = slim.utils.last_dimension(inputs.get_shape(), min_rank=4) - preact = slim.batch_norm(inputs, activation_fn=tf.nn.relu, scope='preact') - if depth == depth_in: - shortcut = subsample(inputs, stride, 'shortcut') - else: - shortcut = slim.conv2d(preact, depth, [1, 1], stride=stride, - normalizer_fn=None, activation_fn=None, - scope='shortcut') - - residual = slim.conv2d(preact, depth_bottleneck, [1, 1], stride=1, - scope='conv1') - residual = conv2d_same(residual, depth_bottleneck, 3, stride, - scope='conv2') - residual = slim.conv2d(residual, depth, [1, 1], stride=1, - normalizer_fn=None, activation_fn=None, - scope='conv3') - - output = shortcut + residual - - return slim.utils.collect_named_outputs(outputs_collections, - sc.name, - output) -# 卷积核权值初始化函数 -def weight_variable(shape): - initial = tf.truncated_normal(shape, stddev=0.1) # 生成一个截断的正态分布，标准差为0.1 - return tf.Variable(initial) - - -# 每个卷积核对应的偏置初始化函数 -def bias_variable(shape): - initial = tf.constant(0.1, shape=shape) - return tf.Variable(initial) - - -def resnet_v2(inputs, - blocks, - embedding_size=None, - global_pool=True, - include_root_block=True, - reuse=None, - scope=None): - """Generator for v2 (preactivation) ResNet models. - This function generates a family of ResNet v2 models. See the resnet_v2_*() - methods for specific model instantiations, obtained by selecting different - block instantiations that produce ResNets of various depths. - Args: - inputs: A tensor of size [batch, height_in, width_in, channels]. - blocks: A list of length equal to the number of ResNet blocks. Each element - is a resnet_utils.Block object describing the units in the block. - num_classes: Number of predicted classes for classification tasks. If None - we return the features before the logit layer. - include_root_block: If True, include the initial convolution followed by - max-pooling, if False excludes it. If excluded, `inputs` should be the - results of an activation-less convolution. - reuse: whether or not the network and its variables should be reused. To be - able to reuse 'scope' must be given. - scope: Optional variable_scope. - Returns: - net: A rank-4 tensor of size [batch, height_out, width_out, channels_out]. - If global_pool is False, then height_out and width_out are reduced by a - factor of output_stride compared to the respective height_in and width_in, - else both height_out and width_out equal one. If num_classes is None, then - net is the output of the last ResNet block, potentially after global - average pooling. If num_classes is not None, net contains the pre-softmax - activations. - end_points: A dictionary from components of the network to the corresponding - activation. - Raises: - ValueError: If the target output_stride is not valid. - """ - with tf.variable_scope(scope, 'resnet_v2', [inputs], reuse=reuse) as sc: - end_points_collection = sc.original_name_scope + '_end_points' - with slim.arg_scope([slim.conv2d, bottleneck, - stack_blocks_dense], - outputs_collections=end_points_collection): - net = inputs - if include_root_block: - # We do not include batch normalization or activation functions in conv1 - # because the first ResNet unit will perform these. Cf. Appendix of [2]. - with slim.arg_scope([slim.conv2d], - activation_fn=None, normalizer_fn=None): - net = conv2d_same(net, 64, 7, stride=2, scope='conv1') - net = slim.max_pool2d(net, [3, 3], stride=2, scope='pool1') - net = stack_blocks_dense(net, blocks) - # This is needed because the pre-activation variant does not have batch - # normalization or activation functions in the residual unit output. See - # Appendix of [2]. - # net = slim.batch_norm(net, activation_fn=tf.nn.relu, scope='postnorm') - # if global_pool: - # # Global average pooling. - # net = tf.reduce_mean(net, [1, 2], name='pool5', keep_dims=True) - # net = slim.batch_norm(net, activation_fn=tf.nn.relu) - - # if num_classes is not None: - # net = slim.conv2d(net, num_classes, [1, 1], activation_fn=None, - # normalizer_fn=None, scope='logits') - # # Convert end_points_collection into a dictionary of end_points. - # end_points = slim.utils.convert_collection_to_dict(end_points_collection) - # if num_classes is not None: - # end_points['predictions'] = slim.softmax(net, scope='predictions') - # - # L.linear(out_dim=self.embedding_dim, name='linear2') - net = tf.reshape(net,(-1,net.shape[1]*net.shape[2]*net.shape[3])) - # net = L.linear(out_dim=512, nl=L.leaky_relu, name='linear1') - W_fc1 = weight_variable([net.shape[1], 512]) - b_fc1 = bias_variable([512]) - # 将中间层与输出层10个结果全连接 - # y_conv = tf.nn.softmax() - ynet = tf.matmul(net, W_fc1) + b_fc1 # [B,512] - ynet = leaky_relu(ynet) - W_fc2 = weight_variable([512,embedding_size]) - b_fc2 = bias_variable([embedding_size]) - - res = tf.matmul(ynet, W_fc2) + b_fc2 # [B,num_classes] - # net = slim.batch_norm(res, activation_fn=tf.nn.relu) - - return res - - -def leaky_relu(x, leak=0.3, name='LeakyRelu'): - """ - leaky_relu - Allow a small non-zero gradient when the unit is not active - Args: - x (tf.tensor): a tensor - leak (float): Default to 0.3 - Returns: - tf.tensor with name 'name' - """ - return tf.maximum(x, leak*x, name=name) - -def resnet_v2_50(inputs, - num_classes=None, - global_pool=True, - reuse=None, - scope='resnet_v2_50'): - """ResNet-50 model of [1]. See resnet_v2() for arg and return description.""" - blocks = [ - Block('block1', bottleneck, [(256, 64, 1)] * 2 + [(256, 64, 2)]), - Block( - 'block2', bottleneck, [(512, 128, 1)] * 3 + [(512, 128, 2)]), - Block( - 'block3', bottleneck, [(1024, 256, 1)] * 5 + [(1024, 256, 2)]), - Block( - 'block4', bottleneck, [(2048, 512, 1)] * 3)] - return resnet_v2(inputs, blocks, num_classes, global_pool, - include_root_block=True, reuse=reuse, scope=scope) - - -def resnet_v2_101(inputs, - num_classes=None, - global_pool=True, - reuse=None, - scope='resnet_v2_101'): - """ResNet-101 model of [1]. See resnet_v2() for arg and return description.""" - blocks = [ - Block( - 'block1', bottleneck, [(256, 64, 1)] * 2 + [(256, 64, 2)]), - Block( - 'block2', bottleneck, [(512, 128, 1)] * 3 + [(512, 128, 2)]), - Block( - 'block3', bottleneck, [(1024, 256, 1)] * 22 + [(1024, 256, 2)]), - Block( - 'block4', bottleneck, [(2048, 512, 1)] * 3)] - return resnet_v2(inputs, blocks, num_classes, global_pool, - include_root_block=True, reuse=reuse, scope=scope) - - -def resnet_v2_152(inputs, - num_classes=None, - global_pool=True, - reuse=None, - scope='resnet_v2_152'): - """ResNet-152 model of [1]. See resnet_v2() for arg and return description.""" - blocks = [ - Block( - 'block1', bottleneck, [(256, 64, 1)] * 2 + [(256, 64, 2)]), - Block( - 'block2', bottleneck, [(512, 128, 1)] * 7 + [(512, 128, 2)]), - Block( - 'block3', bottleneck, [(1024, 256, 1)] * 35 + [(1024, 256, 2)]), - Block( - 'block4', bottleneck, [(2048, 512, 1)] * 3)] - return resnet_v2(inputs, blocks, num_classes, global_pool, - include_root_block=True, reuse=reuse, scope=scope) - - -def resnet_v2_200(inputs, - num_classes=None, - global_pool=True, - reuse=None, - scope='resnet_v2_200'): - """ResNet-200 model of [2]. See resnet_v2() for arg and return description.""" - blocks = [ - Block( - 'block1', bottleneck, [(256, 64, 1)] * 2 + [(256, 64, 2)]), - Block( - 'block2', bottleneck, [(512, 128, 1)] * 23 + [(512, 128, 2)]), - Block( - 'block3', bottleneck, [(1024, 256, 1)] * 35 + [(1024, 256, 2)]), - Block( - 'block4', bottleneck, [(2048, 512, 1)] * 3)] - return resnet_v2(inputs, blocks, num_classes, global_pool, - include_root_block=True, reuse=reuse, scope=scope) - diff --git a/TensorFlow/contrib/cv/CircleLoss_ID1221_for_TensorFlow/models/__pycache__/resnet34.cpython-36.pyc b/TensorFlow/contrib/cv/CircleLoss_ID1221_for_TensorFlow/models/__pycache__/resnet34.cpython-36.pyc new file mode 100644 index 0000000000000000000000000000000000000000..2f4298acf0d90c23c071827a26c5d2b4efe295f8 Binary files /dev/null and b/TensorFlow/contrib/cv/CircleLoss_ID1221_for_TensorFlow/models/__pycache__/resnet34.cpython-36.pyc differ diff --git a/TensorFlow/contrib/cv/CircleLoss_ID1221_for_TensorFlow/models/__pycache__/resnet34.cpython-37.pyc b/TensorFlow/contrib/cv/CircleLoss_ID1221_for_TensorFlow/models/__pycache__/resnet34.cpython-37.pyc new file mode 100644 index 0000000000000000000000000000000000000000..cb89d4b8d66ee71896430ce160af1e43c4cb184f Binary files /dev/null and b/TensorFlow/contrib/cv/CircleLoss_ID1221_for_TensorFlow/models/__pycache__/resnet34.cpython-37.pyc differ diff --git a/TensorFlow/contrib/cv/CircleLoss_ID1221_for_TensorFlow/tool.py b/TensorFlow/contrib/cv/CircleLoss_ID1221_for_TensorFlow/models/resnet34.py similarity index 35% rename from TensorFlow/contrib/cv/CircleLoss_ID1221_for_TensorFlow/tool.py rename to TensorFlow/contrib/cv/CircleLoss_ID1221_for_TensorFlow/models/resnet34.py index b985e2cc95084434e71883c045e7adb19c03b21e..8eb5fafcd93641021875bb5aa7ff4c51f100dabe 100644 --- a/TensorFlow/contrib/cv/CircleLoss_ID1221_for_TensorFlow/tool.py +++ b/TensorFlow/contrib/cv/CircleLoss_ID1221_for_TensorFlow/models/resnet34.py @@ -26,27 +26,62 @@ # See the License for the specific language governing permissions and # limitations under the License. -import os -import shutil - -def copy_code(src,tar): - source_path = os.path.abspath(src) - target_path = os.path.abspath(tar) - - if not os.path.exists(target_path): - os.makedirs(target_path) - - if os.path.exists(source_path): - # root 所指的是当前正在遍历的这个文件夹的本身的地址 - # dirs 是一个 list，内容是该文件夹中所有的目录的名字(不包括子目录) - # files 同样是 list, 内容是该文件夹中所有的文件(不包括子目录) - for root, dirs, files in os.walk(source_path): - if 'logs' in root: - continue - for file in files: - if '.py' in file and '.pyc' not in file or '.sh' in file: - src_file = os.path.join(root, file) - shutil.copy(src_file, target_path) - print(src_file) - - print('copy code finished!') \ No newline at end of file +from npu_bridge.npu_init import * +import tensorflow as tf +import tensorflow.compat.v1 as tf +tf.disable_eager_execution() +import numpy as np + +def inference(images, keep_probability, phase_train=True, + bottleneck_layer_size=128, weight_decay=0.0, reuse=None): + return resnet34(images,phase_train),None + +def bn(x,is_training): + out = tf.layers.batch_normalization(x, training=True,momentum=0.995,epsilon=0.001) + return out + +def conv2d(x,out_channels,kernel_size,strides,padding,activation,useNormalize,is_training): + out=tf.layers.conv2d(inputs=x,filters=out_channels,kernel_size=kernel_size,strides=strides, + padding=padding) + if useNormalize: + out=bn(out,is_training) + if activation: + out=activation(out) + return out + +def BasicBlock(x,is_training,in_channels,out_channels,strides=1): + out = conv2d(x, out_channels, [3, 3], [strides, strides], padding="same", activation=tf.nn.relu, + useNormalize=True,is_training=is_training) + out = conv2d(out, out_channels, [3, 3], [1, 1], padding="same", activation=None, + useNormalize=True,is_training=is_training) + if strides!=1 or in_channels!=out_channels: + x=conv2d(x,out_channels,[1,1],[strides,strides],padding="same",activation=None, + useNormalize=True,is_training=is_training) + out+=x + return tf.nn.relu(out) + +def stage(x,is_training,block,in_channels,out_channels,num_blocks,stride): + strides=[stride]+[1]*(num_blocks-1) + for i in range(len(strides)): + if i==0: + out=block(x,is_training,in_channels,out_channels,strides[i]) + else: + out=block(out,is_training,in_channels,out_channels,strides[i]) + in_channels=out_channels + return out + +def resnet34(x,is_training): + out = conv2d(x, 64, [7, 7], [2, 2], padding="same", activation=tf.nn.relu, + useNormalize=True,is_training=is_training) + #ResNet(BasicBlock, [3,4,6,3]) + out = stage(out,is_training,BasicBlock,64,64,3,1) + out = stage(out,is_training,BasicBlock,64,128,4,2) + out = stage(out,is_training,BasicBlock,128,256,6,2) + out = stage(out,is_training,BasicBlock,256,512,3,2) + out = tf.reduce_mean(out,[1,2],keepdims=True) + size = np.prod(out.get_shape()[1:].as_list()) + out=tf.reshape(out, (-1, size)) + return out + + + diff --git a/TensorFlow/contrib/cv/CircleLoss_ID1221_for_TensorFlow/modelzoo_level.txt b/TensorFlow/contrib/cv/CircleLoss_ID1221_for_TensorFlow/modelzoo_level.txt index 6a684d7ea79aef7e96f28b7153420be46aedaaa1..a093b2ae1acff01ecd8f0f82abbda155062105c7 100644 --- a/TensorFlow/contrib/cv/CircleLoss_ID1221_for_TensorFlow/modelzoo_level.txt +++ b/TensorFlow/contrib/cv/CircleLoss_ID1221_for_TensorFlow/modelzoo_level.txt @@ -1,6 +1,5 @@ GPUStatus:OK NPUMigrationStatus:OK FuncStatus:OK -PrecisionStatus:POK -PerfStatus:OK -PrecisionStatus:POK +PrecisionStatus:OK +PerfStatus:OK \ No newline at end of file diff --git a/TensorFlow/contrib/cv/CircleLoss_ID1221_for_TensorFlow/pip-requirements.txt b/TensorFlow/contrib/cv/CircleLoss_ID1221_for_TensorFlow/pip-requirements.txt deleted file mode 100644 index 7ad9b5a142466959a768887efebc52eeccf3c946..0000000000000000000000000000000000000000 --- a/TensorFlow/contrib/cv/CircleLoss_ID1221_for_TensorFlow/pip-requirements.txt +++ /dev/null @@ -1,5 +0,0 @@ -tensorflow-gpu==1.15 -scikit-image -sklearn -tqdm -tensorboard \ No newline at end of file diff --git a/TensorFlow/contrib/cv/CircleLoss_ID1221_for_TensorFlow/prepare.py b/TensorFlow/contrib/cv/CircleLoss_ID1221_for_TensorFlow/prepare.py deleted file mode 100644 index 74e92ec19eefda5b1bf073b4b178b997179d0b79..0000000000000000000000000000000000000000 --- a/TensorFlow/contrib/cv/CircleLoss_ID1221_for_TensorFlow/prepare.py +++ /dev/null @@ -1,136 +0,0 @@ -# Copyright 2017 The TensorFlow Authors. All Rights Reserved. -# -# Licensed under the Apache License, Version 2.0 (the "License"); -# you may not use this file except in compliance with the License. -# You may obtain a copy of the License at -# -# http://www.apache.org/licenses/LICENSE-2.0 -# -# Unless required by applicable law or agreed to in writing, software -# distributed under the License is distributed on an "AS IS" BASIS, -# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. -# See the License for the specific language governing permissions and -# limitations under the License. -# ============================================================================ -# Copyright 2021 Huawei Technologies Co., Ltd -# -# Licensed under the Apache License, Version 2.0 (the "License"); -# you may not use this file except in compliance with the License. -# You may obtain a copy of the License at -# -# http://www.apache.org/licenses/LICENSE-2.0 -# -# Unless required by applicable law or agreed to in writing, software -# distributed under the License is distributed on an "AS IS" BASIS, -# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. -# See the License for the specific language governing permissions and -# limitations under the License. - -import os -from shutil import copyfile - -# You only need to change this line to your dataset download path -download_path = '/home/nanshen/xutan/yixin/market1501/Market-1501-v15.09.15' - -if not os.path.isdir(download_path): - print('please change the download_path') - -save_path = download_path + '/pytorch' -if not os.path.isdir(save_path): - os.mkdir(save_path) -#----------------------------------------- -#query -query_path = download_path + '/query' -query_save_path = download_path + '/pytorch/query' -if not os.path.isdir(query_save_path): - os.mkdir(query_save_path) - -for root, dirs, files in os.walk(query_path, topdown=True): - for name in files: - if not name[-3:]=='jpg': - continue - ID = name.split('_') - src_path = query_path + '/' + name - dst_path = query_save_path + '/' + ID[0] - if not os.path.isdir(dst_path): - os.mkdir(dst_path) - copyfile(src_path, dst_path + '/' + name) - -#----------------------------------------- -#multi-query -query_path = download_path + '/gt_bbox' -# for dukemtmc-reid, we do not need multi-query -if os.path.isdir(query_path): - query_save_path = download_path + '/pytorch/multi-query' - if not os.path.isdir(query_save_path): - os.mkdir(query_save_path) - - for root, dirs, files in os.walk(query_path, topdown=True): - for name in files: - if not name[-3:]=='jpg': - continue - ID = name.split('_') - src_path = query_path + '/' + name - dst_path = query_save_path + '/' + ID[0] - if not os.path.isdir(dst_path): - os.mkdir(dst_path) - copyfile(src_path, dst_path + '/' + name) - -#----------------------------------------- -#gallery -gallery_path = download_path + '/bounding_box_test' -gallery_save_path = download_path + '/pytorch/gallery' -if not os.path.isdir(gallery_save_path): - os.mkdir(gallery_save_path) - -for root, dirs, files in os.walk(gallery_path, topdown=True): - for name in files: - if not name[-3:]=='jpg': - continue - ID = name.split('_') - src_path = gallery_path + '/' + name - dst_path = gallery_save_path + '/' + ID[0] - if not os.path.isdir(dst_path): - os.mkdir(dst_path) - copyfile(src_path, dst_path + '/' + name) - -#--------------------------------------- -#train_all -train_path = download_path + '/bounding_box_train' -train_save_path = download_path + '/pytorch/train_all' -if not os.path.isdir(train_save_path): - os.mkdir(train_save_path) - -for root, dirs, files in os.walk(train_path, topdown=True): - for name in files: - if not name[-3:]=='jpg': - continue - ID = name.split('_') - src_path = train_path + '/' + name - dst_path = train_save_path + '/' + ID[0] - if not os.path.isdir(dst_path): - os.mkdir(dst_path) - copyfile(src_path, dst_path + '/' + name) - - -#--------------------------------------- -#train_val -train_path = download_path + '/bounding_box_train' -train_save_path = download_path + '/pytorch/train' -val_save_path = download_path + '/pytorch/val' -if not os.path.isdir(train_save_path): - os.mkdir(train_save_path) - os.mkdir(val_save_path) - -for root, dirs, files in os.walk(train_path, topdown=True): - for name in files: - if not name[-3:]=='jpg': - continue - ID = name.split('_') - src_path = train_path + '/' + name - dst_path = train_save_path + '/' + ID[0] - if not os.path.isdir(dst_path): - os.mkdir(dst_path) - dst_path = val_save_path + '/' + ID[0] #first image is used as val image - os.mkdir(dst_path) - copyfile(src_path, dst_path + '/' + name) \ No newline at end of file diff --git a/TensorFlow/contrib/cv/CircleLoss_ID1221_for_TensorFlow/requirements.txt b/TensorFlow/contrib/cv/CircleLoss_ID1221_for_TensorFlow/requirements.txt new file mode 100644 index 0000000000000000000000000000000000000000..871cedac1c338f3ca39e405d6b34a77e6e2a18c9 --- /dev/null +++ b/TensorFlow/contrib/cv/CircleLoss_ID1221_for_TensorFlow/requirements.txt @@ -0,0 +1,35 @@ +absl-py @ file:///opt/conda/conda-bld/absl-py_1639803114343/work +astor==0.8.1 +cached-property @ file:///tmp/build/80754af9/cached-property_1600785575025/work +certifi==2022.5.18.1 +gast==0.2.2 +google-pasta @ file:///Users/ktietz/demo/mc3/conda-bld/google-pasta_1630577991354/work +grpcio==1.14.1 +h5py @ file:///tmp/build/80754af9/h5py_1637138881647/work +imageio==2.19.3 +importlib-metadata @ file:///tmp/build/80754af9/importlib-metadata_1648562407465/work +joblib @ file:///tmp/build/80754af9/joblib_1635411271373/work +Keras-Applications @ file:///tmp/build/80754af9/keras-applications_1594366238411/work +Keras-Preprocessing @ file:///tmp/build/80754af9/keras-preprocessing_1612283640596/work +Markdown @ file:///tmp/build/80754af9/markdown_1614363832606/work +mkl-fft==1.3.1 +mkl-random @ file:///tmp/build/80754af9/mkl_random_1626179032232/work +mkl-service==2.4.0 +numpy @ file:///opt/conda/conda-bld/numpy_and_numpy_base_1651563629415/work +opt-einsum @ file:///tmp/build/80754af9/opt_einsum_1621500238896/work +Pillow==9.1.1 +protobuf==3.20.1 +scikit-learn @ file:///tmp/build/80754af9/scikit-learn_1642601761909/work +scipy @ file:///tmp/build/80754af9/scipy_1641536880743/work +six @ file:///tmp/build/80754af9/six_1644875935023/work +tensorboard==1.15.0 +tensorflow==1.15.0 +tensorflow-estimator==1.15.1 +termcolor==1.1.0 +threadpoolctl @ file:///Users/ktietz/demo/mc3/conda-bld/threadpoolctl_1629802263681/work +tqdm==4.64.0 +typing_extensions @ file:///opt/conda/conda-bld/typing_extensions_1647553014482/work +webencodings==0.5.1 +Werkzeug==0.16.1 +wrapt @ file:///tmp/build/80754af9/wrapt_1638420625940/work +zipp @ file:///opt/conda/conda-bld/zipp_1652341764480/work diff --git a/TensorFlow/contrib/cv/CircleLoss_ID1221_for_TensorFlow/rerank.py b/TensorFlow/contrib/cv/CircleLoss_ID1221_for_TensorFlow/rerank.py deleted file mode 100644 index 43866a0bc5f4e469bcac1c4f1081a168405e91bb..0000000000000000000000000000000000000000 --- a/TensorFlow/contrib/cv/CircleLoss_ID1221_for_TensorFlow/rerank.py +++ /dev/null @@ -1,190 +0,0 @@ -# Copyright 2017 The TensorFlow Authors. All Rights Reserved. -# -# Licensed under the Apache License, Version 2.0 (the "License"); -# you may not use this file except in compliance with the License. -# You may obtain a copy of the License at -# -# http://www.apache.org/licenses/LICENSE-2.0 -# -# Unless required by applicable law or agreed to in writing, software -# distributed under the License is distributed on an "AS IS" BASIS, -# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. -# See the License for the specific language governing permissions and -# limitations under the License. -# ============================================================================ -# Copyright 2021 Huawei Technologies Co., Ltd -# -# Licensed under the Apache License, Version 2.0 (the "License"); -# you may not use this file except in compliance with the License. -# You may obtain a copy of the License at -# -# http://www.apache.org/licenses/LICENSE-2.0 -# -# Unless required by applicable law or agreed to in writing, software -# distributed under the License is distributed on an "AS IS" BASIS, -# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. -# See the License for the specific language governing permissions and -# limitations under the License. - -from ast import parse -import scipy.io -import torch -import numpy as np -import time -# from re_ranking import re_ranking - -import argparse -parser = argparse.ArgumentParser() -parser.add_argument('-p','--path',type=str,default='pytorch_result.mat') -opt = parser.parse_args() - - -import numpy as np - -def k_reciprocal_neigh( initial_rank, i, k1): - forward_k_neigh_index = initial_rank[i,:k1+1] - backward_k_neigh_index = initial_rank[forward_k_neigh_index,:k1+1] - fi = np.where(backward_k_neigh_index==i)[0] - return forward_k_neigh_index[fi] - -def re_ranking(q_g_dist, q_q_dist, g_g_dist, k1=20, k2=6, lambda_value=0.3): - # The following naming, e.g. gallery_num, is different from outer scope. - # Don't care about it. - original_dist = np.concatenate( - [np.concatenate([q_q_dist, q_g_dist], axis=1), - np.concatenate([q_g_dist.T, g_g_dist], axis=1)], - axis=0) - original_dist = 2. - 2 * original_dist # change the cosine similarity metric to euclidean similarity metric - original_dist = np.power(original_dist, 2).astype(np.float32) - original_dist = np.transpose(1. * original_dist/np.max(original_dist,axis = 0)) - V = np.zeros_like(original_dist).astype(np.float32) - #initial_rank = np.argsort(original_dist).astype(np.int32) - # top K1+1 - initial_rank = np.argpartition( original_dist, range(1,k1+1) ) - - query_num = q_g_dist.shape[0] - all_num = original_dist.shape[0] - - for i in range(all_num): - # k-reciprocal neighbors - k_reciprocal_index = k_reciprocal_neigh( initial_rank, i, k1) - k_reciprocal_expansion_index = k_reciprocal_index - for j in range(len(k_reciprocal_index)): - candidate = k_reciprocal_index[j] - candidate_k_reciprocal_index = k_reciprocal_neigh( initial_rank, candidate, int(np.around(k1/2))) - if len(np.intersect1d(candidate_k_reciprocal_index,k_reciprocal_index))> 2./3*len(candidate_k_reciprocal_index): - k_reciprocal_expansion_index = np.append(k_reciprocal_expansion_index,candidate_k_reciprocal_index) - - k_reciprocal_expansion_index = np.unique(k_reciprocal_expansion_index) - weight = np.exp(-original_dist[i,k_reciprocal_expansion_index]) - V[i,k_reciprocal_expansion_index] = 1.*weight/np.sum(weight) - - original_dist = original_dist[:query_num,] - if k2 != 1: - V_qe = np.zeros_like(V,dtype=np.float32) - for i in range(all_num): - V_qe[i,:] = np.mean(V[initial_rank[i,:k2],:],axis=0) - V = V_qe - del V_qe - del initial_rank - invIndex = [] - for i in range(all_num): - invIndex.append(np.where(V[:,i] != 0)[0]) - - jaccard_dist = np.zeros_like(original_dist,dtype = np.float32) - - for i in range(query_num): - temp_min = np.zeros(shape=[1,all_num],dtype=np.float32) - indNonZero = np.where(V[i,:] != 0)[0] - indImages = [] - indImages = [invIndex[ind] for ind in indNonZero] - for j in range(len(indNonZero)): - temp_min[0,indImages[j]] = temp_min[0,indImages[j]]+ np.minimum(V[i,indNonZero[j]],V[indImages[j],indNonZero[j]]) - jaccard_dist[i] = 1-temp_min/(2.-temp_min) - - final_dist = jaccard_dist*(1-lambda_value) + original_dist*lambda_value - del original_dist - del V - del jaccard_dist - final_dist = final_dist[:query_num,query_num:] - return final_dist - -####################################################################### -# Evaluate -def evaluate(score,ql,qc,gl,gc): - index = np.argsort(score) #from small to large - #index = index[::-1] - # good index - query_index = np.argwhere(gl==ql) - camera_index = np.argwhere(gc==qc) - - good_index = np.setdiff1d(query_index, camera_index, assume_unique=True) - junk_index1 = np.argwhere(gl==-1) - junk_index2 = np.intersect1d(query_index, camera_index) - junk_index = np.append(junk_index2, junk_index1) #.flatten()) - - CMC_tmp = compute_mAP(index, good_index, junk_index) - return CMC_tmp - - -def compute_mAP(index, good_index, junk_index): - ap = 0 - cmc = torch.IntTensor(len(index)).zero_() - if good_index.size==0: # if empty - cmc[0] = -1 - return ap,cmc - - # remove junk_index - mask = np.in1d(index, junk_index, invert=True) - index = index[mask] - - # find good_index index - ngood = len(good_index) - mask = np.in1d(index, good_index) - rows_good = np.argwhere(mask==True) - rows_good = rows_good.flatten() - - cmc[rows_good[0]:] = 1 - for i in range(ngood): - d_recall = 1.0/ngood - precision = (i+1)*1.0/(rows_good[i]+1) - if rows_good[i]!=0: - old_precision = i*1.0/rows_good[i] - else: - old_precision=1.0 - ap = ap + d_recall*(old_precision + precision)/2 - - return ap, cmc - -###################################################################### -result = scipy.io.loadmat('/home/liulizhao/projects/liuyixin/projects/checkpoint/pytorch_result.mat') -query_feature = result['query_f'] -query_cam = result['query_cam'][0] -query_label = result['query_label'][0] -gallery_feature = result['gallery_f'] -gallery_cam = result['gallery_cam'][0] -gallery_label = result['gallery_label'][0] - -CMC = torch.IntTensor(len(gallery_label)).zero_() -ap = 0.0 -#re-ranking -print('calculate initial distance') -q_g_dist = np.dot(query_feature, np.transpose(gallery_feature)) -q_q_dist = np.dot(query_feature, np.transpose(query_feature)) -g_g_dist = np.dot(gallery_feature, np.transpose(gallery_feature)) -since = time.time() -re_rank = re_ranking(q_g_dist, q_q_dist, g_g_dist) -time_elapsed = time.time() - since -print('Reranking complete in {:.0f}m {:.0f}s'.format( - time_elapsed // 60, time_elapsed % 60)) -for i in range(len(query_label)): - ap_tmp, CMC_tmp = evaluate(re_rank[i,:],query_label[i],query_cam[i],gallery_label,gallery_cam) - if CMC_tmp[0]==-1: - continue - CMC = CMC + CMC_tmp - ap += ap_tmp - #print(i, CMC_tmp[0]) - -CMC = CMC.float() -CMC = CMC/len(query_label) #average CMC -print('top1:%f top5:%f top10:%f mAP:%f'%(CMC[0],CMC[4],CMC[9],ap/len(query_label))) diff --git a/TensorFlow/contrib/cv/CircleLoss_ID1221_for_TensorFlow/test.py b/TensorFlow/contrib/cv/CircleLoss_ID1221_for_TensorFlow/test.py deleted file mode 100644 index 2038f26ce749d5eab1db2e35693e808d09162134..0000000000000000000000000000000000000000 --- a/TensorFlow/contrib/cv/CircleLoss_ID1221_for_TensorFlow/test.py +++ /dev/null @@ -1,227 +0,0 @@ -# Copyright 2017 The TensorFlow Authors. All Rights Reserved. -# -# Licensed under the Apache License, Version 2.0 (the "License"); -# you may not use this file except in compliance with the License. -# You may obtain a copy of the License at -# -# http://www.apache.org/licenses/LICENSE-2.0 -# -# Unless required by applicable law or agreed to in writing, software -# distributed under the License is distributed on an "AS IS" BASIS, -# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. -# See the License for the specific language governing permissions and -# limitations under the License. -# ============================================================================ -# Copyright 2021 Huawei Technologies Co., Ltd -# -# Licensed under the Apache License, Version 2.0 (the "License"); -# you may not use this file except in compliance with the License. -# You may obtain a copy of the License at -# -# http://www.apache.org/licenses/LICENSE-2.0 -# -# Unless required by applicable law or agreed to in writing, software -# distributed under the License is distributed on an "AS IS" BASIS, -# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. -# See the License for the specific language governing permissions and -# limitations under the License. - -from __future__ import print_function, division - -import argparse -import torch -import torch.nn as nn -from torch.autograd import Variable -import torch.backends.cudnn as cudnn -import numpy as np -from torchvision import datasets, models, transforms -import os -import scipy.io -import math -# os.system('cd /home/nanshen/xutan/yixin/circlesoftmax') -parser = argparse.ArgumentParser(description='Test') -parser.add_argument('--graphpath',default='/home/nanshen/xutan/yixin/circlesoftmax/logs/b16_epoch10_lr5en2/ckpt/5656/softmaxcircleloss.ckpt-5656.meta', type=str) -parser.add_argument('--parampath',default='/home/nanshen/xutan/yixin/circlesoftmax/logs/b16_epoch10_lr5en2/ckpt/5656', type=str) -parser.add_argument('--logdir',default='/home/nanshen/xutan/yixin/circlesoftmax/logs/b16_epoch10_lr5en2', type=str) -parser.add_argument('--gpu_ids',default='0', type=str,help='gpu_ids: e.g. 0 0,1,2 0,2') -parser.add_argument('--test_dir',default='Data/Market-1501-v15.09.15/pytorch',type=str, help='./test_data') -parser.add_argument('--savepath', default='checkpoint', type=str, help='save model path') -parser.add_argument('--batchsize', default=32, type=int, help='batchsize') -parser.add_argument('--multi', action='store_true', help='use multiple query' ) -parser.add_argument('--ms',default='1', type=str,help='multiple_scale: e.g. 1 1,1.1 1,1.1,1.2') - -opt = parser.parse_args() - -opt.nclasses = 751 - -str_ids = opt.gpu_ids.split(',') -#which_epoch = opt.which_epoch -# name = opt.name -test_dir = opt.test_dir - -gpu_ids = [] -for str_id in str_ids: - id = int(str_id) - if id >=0: - gpu_ids.append(id) - -print('We use the scale: %s'%opt.ms) -str_ms = opt.ms.split(',') -ms = [] -for s in str_ms: - s_f = float(s) - ms.append(math.sqrt(s_f)) - -# set gpu ids -# if len(gpu_ids)>0: -# torch.cuda.set_device(1) -# cudnn.benchmark = True - - -### tf model ##### -import os -os.environ["CUDA_VISIBLE_DEVICES"]=opt.gpu_ids -import tensorflow as tf - -sess=tf.Session() -saver = tf.train.import_meta_graph(opt.graphpath) -saver.restore(sess, tf.train.latest_checkpoint(opt.parampath)) - -# 访问placeholders变量，并且创建feed-dict来作为placeholders的新值 -graph = tf.get_default_graph() - -x = graph.get_tensor_by_name("inputs:0") -keep_rate = graph.get_tensor_by_name("keep_rate:0") -# y = graph.get_tensor_by_name("labels:0") -istrain = graph.get_tensor_by_name("is_training:0") -features = graph.get_tensor_by_name("Logits/features:0") -def extract_feature_tf(batch_im): - return sess.run(features,feed_dict={x:batch_im,istrain:False,keep_rate:1.0}) - -###################################################################### -# Load Data -# --------- -# -# We will use torchvision and torch.utils.data packages for loading the -# data. -# -h, w = 256, 128 - - -data_transforms = transforms.Compose([ - transforms.Resize((h, w), interpolation=3), - transforms.ToTensor(), - transforms.Normalize([0.485, 0.456, 0.406], [0.229, 0.224, 0.225]) -############### Ten Crop - #transforms.TenCrop(224), - #transforms.Lambda(lambda crops: torch.stack( - # [transforms.ToTensor()(crop) - # for crop in crops] - # )), - #transforms.Lambda(lambda crops: torch.stack( - # [transforms.Normalize([0.485, 0.456, 0.406], [0.229, 0.224, 0.225])(crop) - # for crop in crops] - # )) -]) - -data_dir = test_dir - -if opt.multi: - image_datasets = {x: datasets.ImageFolder( os.path.join(data_dir,x) ,data_transforms) for x in ['gallery','query','multi-query']} - dataloaders = {x: torch.utils.data.DataLoader(image_datasets[x], batch_size=opt.batchsize, - shuffle=False, num_workers=16) for x in ['gallery','query','multi-query']} -else: - image_datasets = {x: datasets.ImageFolder( os.path.join(data_dir,x) ,data_transforms) for x in ['gallery','query']} - dataloaders = {x: torch.utils.data.DataLoader(image_datasets[x], batch_size=opt.batchsize, - shuffle=False, num_workers=16) for x in ['gallery','query']} -class_names = image_datasets['query'].classes -# use_gpu = torch.cuda.is_available() - - - -###################################################################### -# Extract feature -# ---------------------- -# -# Extract feature from a trained model. -# -def fliplr(img): - '''flip horizontal''' - inv_idx = torch.arange(img.size(3)-1,-1,-1).long() # N x C x H x W - img_flip = img.index_select(3,inv_idx) - return img_flip - -def extract_feature(dataloaders): - features = torch.FloatTensor() - count = 0 - from tqdm import tqdm - for data in tqdm(dataloaders): - img, label = data - n, c, h, w = img.size() - count += n - # print(count) - ff = torch.FloatTensor(n,512).zero_() - for i in range(2): - if(i==1): - img = fliplr(img) - # input_img = Variable(img.cuda()) - input_img = img.cpu().numpy() - for scale in ms: - if scale != 1: - # bicubic is only available in pytorch>= 1.1 - input_img = nn.functional.interpolate(torch.tensor(input_img), scale_factor=scale, mode='bicubic', align_corners=False) - input_img = input_img.cpu().numpy() - # outputs = extract_feature_tf(input_img) - input_img = input_img.transpose(0,2,3,1) - outputs = extract_feature_tf(input_img) - outputs = Variable(torch.from_numpy(outputs)) - ff += outputs - features = torch.cat((features,ff.data.cpu()), 0) - return features - -def get_id(img_path): - camera_id = [] - labels = [] - for path, v in img_path: - #filename = path.split('/')[-1] - filename = os.path.basename(path) - label = filename[0:4] - camera = filename.split('c')[1] - if label[0:2]=='-1': - labels.append(-1) - else: - labels.append(int(label)) - camera_id.append(int(camera[0])) - return camera_id, labels - -gallery_path = image_datasets['gallery'].imgs -query_path = image_datasets['query'].imgs - -gallery_cam,gallery_label = get_id(gallery_path) -query_cam,query_label = get_id(query_path) - -if opt.multi: - mquery_path = image_datasets['multi-query'].imgs - mquery_cam,mquery_label = get_id(mquery_path) - - -# Extract feature -gallery_feature = extract_feature(dataloaders['gallery']) -query_feature = extract_feature(dataloaders['query']) -if opt.multi: - mquery_feature = extract_feature(dataloaders['multi-query']) - -import time -timestamp = int(time.time()) - -# Save to Matlab for check -result = {'gallery_f':gallery_feature.numpy(),'gallery_label':gallery_label,'gallery_cam':gallery_cam,'query_f':query_feature.numpy(),'query_label':query_label,'query_cam':query_cam} -scipy.io.savemat(f'res/pytorch_result_{timestamp}.mat',result) - -# print(opt.name) -# result = 'checkpoint/result.txt' -os.system(f'python code/eval.py --filename res/pytorch_result_{timestamp}.mat --logdir {opt.logdir}') - -if opt.multi: - result = {'mquery_f':mquery_feature.numpy(),'mquery_label':mquery_label,'mquery_cam':mquery_cam} - scipy.io.savemat(opt.savepath+'multi_query.mat',result) diff --git a/TensorFlow/contrib/cv/CircleLoss_ID1221_for_TensorFlow/test/train_full_1p.sh b/TensorFlow/contrib/cv/CircleLoss_ID1221_for_TensorFlow/test/train_full_1p.sh deleted file mode 100644 index c8863f2608eef9ff2f6c1b869f4f01e357ee6fb9..0000000000000000000000000000000000000000 --- a/TensorFlow/contrib/cv/CircleLoss_ID1221_for_TensorFlow/test/train_full_1p.sh +++ /dev/null @@ -1,103 +0,0 @@ -#!/bin/bash - - -export RANK_SIZE=1 -export JOB_ID=10087 -export RANK_ID_START=0 - - -cur_path=`pwd` -data_path="" -ckpt_path="" -Network="CircleLoss_ID1221_for_TensorFlow" -batch_size=16 -epochs=120 -# train_performance_1p.sh perf -# train_full_1p.sh acc -CaseName="${Network}_bs${batch_size}_${RANK_SIZE}p_acc" - - -if [[ $1 == --help || $1 == -h ]];then - echo"usage:./train_full_1p.sh " - echo " " - echo "parameter explain: - --precision_mode precision mode(allow_fp32_to_fp16/force_fp16/must_keep_origin_dtype/allow_mix_precision) - --over_dump if or not over detection, default is False - --data_dump_flag data dump flag, default is False - --data_dump_step data dump step, default is 10 - --profiling if or not profiling for performance debug, default is False - --data_path source data of training - -h/--help show help message - " - exit 1 -fi -for para in $* -do - if [[ $para == --data_path* ]];then - data_path=`echo ${para#*=}` - echo "${data_path}" - elif [[ $para == --ckpt_path* ]];then - ckpt_path=`echo ${para#*=}` - echo "${ckpt_path}" - elif [[ $para == --batch_size* ]];then - batch_size=`echo ${para#*=}` - echo "${batch_size}" - elif [[ $para == --epochs* ]];then - epochs=`echo ${para#*=}` - echo "${epochs}" - fi -done -if [[ $data_path == "" ]];then - echo "[Error] para \"data_path\" must be confing" - exit 1 -fi - - -cd $cur_path/../ -# START -start_time=$(date +%s) -for((RANK_ID=$RANK_ID_START;RANK_ID<$((RANK_SIZE+RANK_ID_START));RANK_ID++)); -do - echo "Device ID: $ASCEND_DEVICE_ID" - export RANK_ID=$RANK_ID - if [ -d ${cur_path}/output/${ASCEND_DEVICE_ID} ];then - rm -rf ${cur_path}/output/${ASCEND_DEVICE_ID} - mkdir -p ${cur_path}/output/${ASCEND_DEVICE_ID}/ckpt - else - mkdir -p ${cur_path}/output/${ASCEND_DEVICE_ID}/ckpt - fi - nohup python3 -u train_npuv3.py --data_path=${data_path} --epoch_size=${epochs} > ${cur_path}/output/${ASCEND_DEVICE_ID}/train_${ASCEND_DEVICE_ID}.log 2>&1 & -done -wait -end_time=$(date +%s) -e2e_time=$(( ${end_time} - ${start_time} )) - - -echo "------------------ Final result ------------------" -BatchSize=${batch_size} -DeviceType=`uname -m` -# getFPS -TimePerStep=`grep 'Epoch' ${cur_path}/output/${ASCEND_DEVICE_ID}/train_${ASCEND_DEVICE_ID}.log | awk -F'TimePerStep:' '{print $2}' | awk -F's' 'END{print $1}'` -FPS=`awk 'BEGIN{printf "%.2f\n",'${batch_size}'/'${TimePerStep}'}'` -ActualFPS=${FPS} -TrainingTime=`awk 'BEGIN{printf "%.2f\n",'${BatchSize}'/'${FPS}'}'` -# getAcc -train_accuracy="None" -# getLoss -grep 'Epoch' ${cur_path}/output/${ASCEND_DEVICE_ID}/train_${ASCEND_DEVICE_ID}.log | awk -F'TrainLoss:' '{print $2}' | awk -F' ' '{print $1}' > ${cur_path}/output/${ASCEND_DEVICE_ID}/train_${CaseName}_loss.txt -ActualLoss=`awk 'END {print}' ${cur_path}/output/${ASCEND_DEVICE_ID}/train_${CaseName}_loss.txt` -echo "Final Performance images/sec : ${FPS}" -echo "Final Train Accuracy : ${train_accuracy}" -echo "E2E Training Duration sec : ${e2e_time}" - - -echo "Network = ${Network}" > ${cur_path}/output/${ASCEND_DEVICE_ID}/${CaseName}.log -echo "RankSize = ${RANK_SIZE}" >> ${cur_path}/output/${ASCEND_DEVICE_ID}/${CaseName}.log -echo "BatchSize = ${BatchSize}" >> ${cur_path}/output/${ASCEND_DEVICE_ID}/${CaseName}.log -echo "DeviceType = ${DeviceType}" >> ${cur_path}/output/${ASCEND_DEVICE_ID}/${CaseName}.log -echo "CaseName = ${CaseName}" >> ${cur_path}/output/${ASCEND_DEVICE_ID}/${CaseName}.log -echo "ActualFPS = ${ActualFPS}" >> ${cur_path}/output/${ASCEND_DEVICE_ID}/${CaseName}.log -echo "TrainingTime = ${TrainingTime}" >> ${cur_path}/output/${ASCEND_DEVICE_ID}/${CaseName}.log -echo "TrainAccuracy = ${train_accuracy}" >> ${cur_path}/output/${ASCEND_DEVICE_ID}/${CaseName}.log -echo "ActualLoss = ${ActualLoss}" >> ${cur_path}/output/${ASCEND_DEVICE_ID}/${CaseName}.log -echo "E2ETrainingTime = ${e2e_time}" >> ${cur_path}/output/${ASCEND_DEVICE_ID}/${CaseName}.log diff --git a/TensorFlow/contrib/cv/CircleLoss_ID1221_for_TensorFlow/test/train_performance_1p.sh b/TensorFlow/contrib/cv/CircleLoss_ID1221_for_TensorFlow/test/train_performance_1p.sh deleted file mode 100644 index c049f0dd450acddffde7f23e4d0ffbf73d81ed2c..0000000000000000000000000000000000000000 --- a/TensorFlow/contrib/cv/CircleLoss_ID1221_for_TensorFlow/test/train_performance_1p.sh +++ /dev/null @@ -1,103 +0,0 @@ -#!/bin/bash - - -export RANK_SIZE=1 -export JOB_ID=10087 -export RANK_ID_START=0 - - -cur_path=`pwd` -data_path="" -ckpt_path="" -Network="CircleLoss_ID1221_for_TensorFlow" -batch_size=16 -epochs=1 -# train_performance_1p.sh perf -# train_full_1p.sh acc -CaseName="${Network}_bs${batch_size}_${RANK_SIZE}p_perf" - - -if [[ $1 == --help || $1 == -h ]];then - echo"usage:./train_performance_1p.sh " - echo " " - echo "parameter explain: - --precision_mode precision mode(allow_fp32_to_fp16/force_fp16/must_keep_origin_dtype/allow_mix_precision) - --over_dump if or not over detection, default is False - --data_dump_flag data dump flag, default is False - --data_dump_step data dump step, default is 10 - --profiling if or not profiling for performance debug, default is False - --data_path source data of training - -h/--help show help message - " - exit 1 -fi -for para in $* -do - if [[ $para == --data_path* ]];then - data_path=`echo ${para#*=}` - echo "${data_path}" - elif [[ $para == --ckpt_path* ]];then - ckpt_path=`echo ${para#*=}` - echo "${ckpt_path}" - elif [[ $para == --batch_size* ]];then - batch_size=`echo ${para#*=}` - echo "${batch_size}" - elif [[ $para == --epochs* ]];then - epochs=`echo ${para#*=}` - echo "${epochs}" - fi -done -if [[ $data_path == "" ]];then - echo "[Error] para \"data_path\" must be confing" - exit 1 -fi - - -cd $cur_path/../ -# START -start_time=$(date +%s) -for((RANK_ID=$RANK_ID_START;RANK_ID<$((RANK_SIZE+RANK_ID_START));RANK_ID++)); -do - echo "Device ID: $ASCEND_DEVICE_ID" - export RANK_ID=$RANK_ID - if [ -d ${cur_path}/output/${ASCEND_DEVICE_ID} ];then - rm -rf ${cur_path}/output/${ASCEND_DEVICE_ID} - mkdir -p ${cur_path}/output/${ASCEND_DEVICE_ID}/ckpt - else - mkdir -p ${cur_path}/output/${ASCEND_DEVICE_ID}/ckpt - fi - nohup python3 -u train_npuv3.py --data_path=${data_path} --epoch_size=${epochs} > ${cur_path}/output/${ASCEND_DEVICE_ID}/train_${ASCEND_DEVICE_ID}.log 2>&1 & -done -wait -end_time=$(date +%s) -e2e_time=$(( ${end_time} - ${start_time} )) - - -echo "------------------ Final result ------------------" -BatchSize=${batch_size} -DeviceType=`uname -m` -# getFPS -TimePerStep=`grep 'Epoch' ${cur_path}/output/${ASCEND_DEVICE_ID}/train_${ASCEND_DEVICE_ID}.log | awk -F'TimePerStep:' '{print $2}' | awk -F's' 'END{print $1}'` -FPS=`awk 'BEGIN{printf "%.2f\n",'${batch_size}'/'${TimePerStep}'}'` -ActualFPS=${FPS} -TrainingTime=`awk 'BEGIN{printf "%.2f\n",'${BatchSize}'/'${FPS}'}'` -# getAcc -train_accuracy="None" -# getLoss -grep 'Epoch' ${cur_path}/output/${ASCEND_DEVICE_ID}/train_${ASCEND_DEVICE_ID}.log | awk -F'TrainLoss:' '{print $2}' | awk -F' ' '{print $1}' > ${cur_path}/output/${ASCEND_DEVICE_ID}/train_${CaseName}_loss.txt -ActualLoss=`awk 'END {print}' ${cur_path}/output/${ASCEND_DEVICE_ID}/train_${CaseName}_loss.txt` -echo "Final Performance images/sec : ${FPS}" -echo "Final Train Accuracy : ${train_accuracy}" -echo "E2E Training Duration sec : ${e2e_time}" - - -echo "Network = ${Network}" > ${cur_path}/output/${ASCEND_DEVICE_ID}/${CaseName}.log -echo "RankSize = ${RANK_SIZE}" >> ${cur_path}/output/${ASCEND_DEVICE_ID}/${CaseName}.log -echo "BatchSize = ${BatchSize}" >> ${cur_path}/output/${ASCEND_DEVICE_ID}/${CaseName}.log -echo "DeviceType = ${DeviceType}" >> ${cur_path}/output/${ASCEND_DEVICE_ID}/${CaseName}.log -echo "CaseName = ${CaseName}" >> ${cur_path}/output/${ASCEND_DEVICE_ID}/${CaseName}.log -echo "ActualFPS = ${ActualFPS}" >> ${cur_path}/output/${ASCEND_DEVICE_ID}/${CaseName}.log -echo "TrainingTime = ${TrainingTime}" >> ${cur_path}/output/${ASCEND_DEVICE_ID}/${CaseName}.log -echo "TrainAccuracy = ${train_accuracy}" >> ${cur_path}/output/${ASCEND_DEVICE_ID}/${CaseName}.log -echo "ActualLoss = ${ActualLoss}" >> ${cur_path}/output/${ASCEND_DEVICE_ID}/${CaseName}.log -echo "E2ETrainingTime = ${e2e_time}" >> ${cur_path}/output/${ASCEND_DEVICE_ID}/${CaseName}.log diff --git a/TensorFlow/contrib/cv/CircleLoss_ID1221_for_TensorFlow/test/train_testcase.sh b/TensorFlow/contrib/cv/CircleLoss_ID1221_for_TensorFlow/test/train_testcase.sh deleted file mode 100644 index faf25281e347fe6c690603affdf0da2c4553742e..0000000000000000000000000000000000000000 --- a/TensorFlow/contrib/cv/CircleLoss_ID1221_for_TensorFlow/test/train_testcase.sh +++ /dev/null @@ -1,31 +0,0 @@ -#! /bin/bash -#开发者个人独立预置的数据集、预训练模型、ATC-OM模型等，支持从OBS仓下载 -obsutil cp obs://liuyixin2/datasets/Market-1501-v15.09.15/resnet_v1_50.ckpt ./dataset/ -f -r -obsutil cp obs://liuyixin2/datasets/Market-1501-v15.09.15/bounding_box_train.zip ./dataset/ -f -r -unzip -d ./dataset/ ./dataset/bounding_box_train.zip -pip3 install -r requirements.txt - - -#testcase主体，开发者根据不同模型写作 -batch_size=16 -epochs=1 -data_url="./dataset/" -train_url="./dataset/" - -nohup python3.7 train_npuv3.py \ - --data_url="${data_url}" \ - --epoch_size="${epochs}" \ - --train_url="${train_url}" > train.log 2>&1 - -wait - -#结果判断，功能检查输出ckpt/日志关键字、精度检查loss值/accucy关键字、性能检查耗时打点/ThroughOutput等关键字 -key1="Train success" #功能检查字 -key2="images/sec" #性能检查字 -key3="log10" #精度检查字 - -if [ `grep -c "$key1" "tran.log"` -ne '0' ] ;then #可以根据需要调整检查逻辑 - echo "Run testcase success!" -else - echo "Run testcase failed!" -fi \ No newline at end of file diff --git a/TensorFlow/contrib/cv/CircleLoss_ID1221_for_TensorFlow/test_gpu.py b/TensorFlow/contrib/cv/CircleLoss_ID1221_for_TensorFlow/test_gpu.py new file mode 100644 index 0000000000000000000000000000000000000000..193533bd65e95c81e9241adfa538d832eca13333 --- /dev/null +++ b/TensorFlow/contrib/cv/CircleLoss_ID1221_for_TensorFlow/test_gpu.py @@ -0,0 +1,521 @@ +# Copyright 2017 The TensorFlow Authors. All Rights Reserved. +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. +# ============================================================================ +# Copyright 2021 Huawei Technologies Co., Ltd +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. +from __future__ import absolute_import +from __future__ import division +from __future__ import print_function + +from datetime import datetime +import os.path +import time +import sys +import random +import tensorflow.compat.v1 as tf +tf.disable_eager_execution() +import numpy as np +import importlib +import argparse +import facenet +import lfw +import h5py +import math +from circle_loss import * +from facenet import load_data +from PIL import Image + +def fully_connected(x, units, activation_fn=None,reuse=False): + out=tf.layers.dense(inputs=x,units=units,activation=activation_fn,reuse=reuse) + return out + +def load_and_preprocess_image(img_path): + img_raw = tf.io.read_file(img_path) + img_tensor = tf.cond( + tf.image.is_jpeg(img_raw), + lambda: tf.image.decode_jpeg(img_raw,3), + lambda: tf.image.decode_png(img_raw,3)) + img_tensor = tf.image.resize(img_tensor, [112, 112]) + img_tensor = tf.cast(img_tensor, tf.float32) + img_tensor = (img_tensor - 127.5) / 128.0 + return img_tensor + +def getIteration(args, image_paths_placeholder, labels_placeholder): + image_paths_dataset=tf.data.Dataset.from_tensor_slices(image_paths_placeholder).map(load_and_preprocess_image) + label_dataset=tf.data.Dataset.from_tensor_slices(labels_placeholder) + dataset=tf.data.Dataset.zip((image_paths_dataset,label_dataset)).prefetch( + args.batch_size).batch(args.batch_size) + iteration=dataset.make_initializable_iterator() + return iteration + +def main(args): + network = importlib.import_module(args.model_def) + subdir = datetime.strftime(datetime.now(), '%Y%m%d-%H%M%S') + log_dir = os.path.join(os.path.expanduser(args.logs_base_dir), subdir) + + if not os.path.isdir(log_dir): + os.makedirs(log_dir) + model_dir = os.path.join(os.path.expanduser(args.models_base_dir), subdir) + if not os.path.isdir(model_dir): + os.makedirs(model_dir) + + stat_file_name = os.path.join(log_dir, 'stat.h5') + + facenet.write_arguments_to_file(args, os.path.join(log_dir, 'arguments.txt')) + + src_path, _ = os.path.split(os.path.realpath(__file__)) + facenet.store_revision_info(src_path, log_dir, ' '.join(sys.argv)) + + np.random.seed(seed=args.seed) + random.seed(args.seed) + dataset = facenet.get_dataset(args.data_dir) + if args.filter_filename: + dataset = filter_dataset(dataset, os.path.expanduser(args.filter_filename), + args.filter_percentile, args.filter_min_nrof_images_per_class) + + if args.validation_set_split_ratio > 0.0: + train_set, val_set = facenet.split_dataset(dataset, args.validation_set_split_ratio, + args.min_nrof_val_images_per_class, 'SPLIT_IMAGES') + else: + train_set, val_set = dataset, [] + + nrof_classes = len(train_set) + + print('Model directory: %s' % model_dir) + #print('Log directory: %s' % log_dir) + pretrained_model = None + if args.pretrained_model: + pretrained_model = os.path.expanduser(args.pretrained_model) + print('Pre-trained model: %s' % pretrained_model) + + if args.lfw_dir: + print('LFW directory: %s' % args.lfw_dir) + pairs = lfw.read_pairs(os.path.expanduser(args.lfw_pairs)) + lfw_paths, actual_issame = lfw.get_paths(os.path.expanduser(args.lfw_dir), pairs) + + with tf.Graph().as_default(): + tf.set_random_seed(args.seed) + global_step = tf.Variable(0, trainable=False) + + image_list, label_list = facenet.get_image_paths_and_labels(train_set) + assert len(image_list) > 0, 'The training set should not be empty' + + val_image_list, val_label_list = facenet.get_image_paths_and_labels(val_set) + + learning_rate_placeholder = tf.placeholder(tf.float32, name='learning_rate') + phase_train_placeholder = tf.placeholder(tf.bool, name='phase_train') + image_paths_placeholder=tf.placeholder(tf.string,shape=(None,),name="image_paths") + labels_placeholder=tf.placeholder(tf.int32,shape=(None,),name="labels") + + iteration = getIteration(args, image_paths_placeholder, labels_placeholder) + image_batch,label_batch=iteration.get_next() + + image_batch = tf.identity(image_batch, 'image_batch') + image_batch = tf.identity(image_batch, 'input') + label_batch = tf.identity(label_batch, 'label_batch') + + # print('Number of classes in training set: %d' % nrof_classes) + # print('Number of examples in training set: %d' % len(image_list)) + # + # print('Number of classes in validation set: %d' % len(val_set)) + # print('Number of examples in validation set: %d' % len(val_image_list)) + + # print('Building training graph') + + prelogits, _ = network.inference(image_batch, args.keep_probability, + phase_train=phase_train_placeholder, bottleneck_layer_size=args.embedding_size, + weight_decay=args.weight_decay) + logits=fully_connected(prelogits,len(train_set),None,False) + + norm_feat = tf.nn.l2_normalize(logits, 1, 1e-10, name='norm_feat') + + circle_loss_mean = circle_loss(tf.one_hot(label_batch, len(train_set), dtype=tf.float32),norm_feat,batch_size=args.batch_size) + + embeddings = tf.nn.l2_normalize(prelogits, 1, 1e-10, name='embeddings') + + + eps = 1e-4 + prelogits_norm = tf.reduce_mean(tf.norm(tf.abs(prelogits) + eps, ord=args.prelogits_norm_p, axis=1)) + tf.add_to_collection(tf.GraphKeys.REGULARIZATION_LOSSES, prelogits_norm * args.prelogits_norm_loss_factor) + + prelogits_center_loss, _ = facenet.center_loss(prelogits, label_batch, args.center_loss_alfa, nrof_classes) + tf.add_to_collection(tf.GraphKeys.REGULARIZATION_LOSSES, prelogits_center_loss * args.center_loss_factor) + + learning_rate = tf.train.exponential_decay(learning_rate_placeholder, global_step, + args.learning_rate_decay_epochs * args.epoch_size, + args.learning_rate_decay_factor, staircase=True) + tf.summary.scalar('learning_rate', learning_rate) + + tf.add_to_collection('losses', circle_loss_mean) + + correct_prediction = tf.cast(tf.equal(tf.argmax(logits, 1), tf.cast(label_batch, tf.int64)), tf.float32) + accuracy = tf.reduce_mean(correct_prediction) + + regularization_losses = tf.get_collection(tf.GraphKeys.REGULARIZATION_LOSSES) + + total_loss = tf.add_n([circle_loss_mean] + regularization_losses, name='total_loss') + + train_op = facenet.train(total_loss, global_step, args.optimizer, + learning_rate, args.moving_average_decay, tf.global_variables(), args.log_histograms) + + saver = tf.train.Saver(tf.trainable_variables(), max_to_keep=3) + + summary_op = tf.summary.merge_all() + + gpu_options = tf.GPUOptions(per_process_gpu_memory_fraction=args.gpu_memory_fraction) + + sess = tf.Session(config=tf.ConfigProto(gpu_options=gpu_options, log_device_placement=False)) + sess.run(tf.global_variables_initializer()) + sess.run(tf.local_variables_initializer()) + + summary_writer = tf.summary.FileWriter(log_dir, sess.graph) + coord = tf.train.Coordinator() + tf.train.start_queue_runners(coord=coord, sess=sess) + + with sess.as_default(): + + if pretrained_model: + print('Restoring pretrained model: %s' % pretrained_model) + saver.restore(sess, pretrained_model) + + # print('Running training') + nrof_steps = args.max_nrof_epochs * args.epoch_size + nrof_val_samples = int(math.ceil( + args.max_nrof_epochs / args.validate_every_n_epochs)) + stat = { + 'loss': np.zeros((nrof_steps,), np.float32), + 'center_loss': np.zeros((nrof_steps,), np.float32), + 'reg_loss': np.zeros((nrof_steps,), np.float32), + 'xent_loss': np.zeros((nrof_steps,), np.float32), + 'prelogits_norm': np.zeros((nrof_steps,), np.float32), + 'accuracy': np.zeros((nrof_steps,), np.float32), + 'val_loss': np.zeros((nrof_val_samples,), np.float32), + 'val_xent_loss': np.zeros((nrof_val_samples,), np.float32), + 'val_accuracy': np.zeros((nrof_val_samples,), np.float32), + 'lfw_accuracy': np.zeros((args.max_nrof_epochs,), np.float32), + 'lfw_valrate': np.zeros((args.max_nrof_epochs,), np.float32), + 'learning_rate': np.zeros((args.max_nrof_epochs,), np.float32), + 'time_train': np.zeros((args.max_nrof_epochs,), np.float32), + 'time_validate': np.zeros((args.max_nrof_epochs,), np.float32), + 'time_evaluate': np.zeros((args.max_nrof_epochs,), np.float32), + 'prelogits_hist': np.zeros((args.max_nrof_epochs, 1000), np.float32), + } + + for epoch in range(1, args.max_nrof_epochs + 1): + step = sess.run(global_step, feed_dict=None) + + save_variables_and_metagraph(sess, saver, summary_writer, model_dir, subdir, epoch) + + t = time.time() + if args.lfw_dir: + acc_mean,acc_var=evaluate(sess,iteration,lfw_paths, image_paths_placeholder, labels_placeholder, phase_train_placeholder, + embeddings,label_batch,actual_issame, args.lfw_batch_size, + args.lfw_nrof_folds, log_dir, step, summary_writer,stat,epoch, + args.lfw_distance_metric, args.lfw_subtract_mean, args.lfw_use_flipped_images, + args.use_fixed_image_standardization) + fout=open("result.txt","a+") + fout.write(str(epoch)+","+str(acc_mean)+"+-"+str(acc_var)+"\n") + fout.close() + stat['time_evaluate'][epoch - 1] = time.time() - t + + # print('Saving statistics') + with h5py.File(stat_file_name, 'w') as f: + for key, value in stat.items(): + f.create_dataset(key, data=value) + + return model_dir + +def train_epoch(args,sess,epoch,iteration,image_list,label_list,image_paths_placeholder,labels_placeholder, + learning_rate_placeholder,phase_train_placeholder,step, + loss,train_op,summary_op,summary_writer,reg_losses,learning_rate_schedule_file, + stat,circle_loss_mean,accuracy, + learning_rate,prelogits,prelogits_center_loss,random_rotate,random_crop,random_flip,prelogits_norm, + prelogits_hist_max,use_fixed_image_standardization): + + epoch_idx=np.array(random.sample(range(len(label_list)),args.epoch_size*args.batch_size)) + image_list=np.array(image_list)[epoch_idx] + label_list=np.array(label_list)[epoch_idx] + + batch_number = 0 + if args.learning_rate > 0.0: + lr = args.learning_rate + else: + lr = facenet.get_learning_rate_from_file(learning_rate_schedule_file, epoch) + + if lr <= 0: + return False + feed_dict={image_paths_placeholder:image_list, + labels_placeholder:label_list} + sess.run(iteration.initializer,feed_dict=feed_dict) + + train_time = 0 + while batch_number < args.epoch_size: + start_time = time.time() + + feed_dict = {learning_rate_placeholder: lr, + phase_train_placeholder: True} + + tensor_list = [loss, train_op, step, reg_losses, prelogits, circle_loss_mean, learning_rate, prelogits_norm, + accuracy, prelogits_center_loss] + if batch_number % 100 == 0: + loss_, _, step_, reg_losses_, prelogits_, circle_loss_mean_, lr_, prelogits_norm_, accuracy_, center_loss_, summary_str = sess.run(tensor_list + [summary_op], feed_dict=feed_dict) + summary_writer.add_summary(summary_str, global_step=step_) + else: + loss_, _, step_, reg_losses_, prelogits_, circle_loss_mean_, lr_, prelogits_norm_, accuracy_, center_loss_ = sess.run(tensor_list, feed_dict=feed_dict) + + stat['loss'][step_ - 1] = loss_ + stat['center_loss'][step_ - 1] = center_loss_ + stat['reg_loss'][step_ - 1] = np.sum(reg_losses_) + stat['xent_loss'][step_ - 1] = circle_loss_mean_ + stat['prelogits_norm'][step_ - 1] = prelogits_norm_ + stat['learning_rate'][epoch - 1] = lr_ + stat['accuracy'][step_ - 1] = accuracy_ + stat['prelogits_hist'][epoch - 1, :] += \ + np.histogram(np.minimum(np.abs(prelogits_), prelogits_hist_max), bins=1000, range=(0.0, prelogits_hist_max))[0] + + duration = time.time() - start_time + # print( + # 'Epoch: [%d][%d/%d]\tTime %.3f\tLoss %2.3f\tXent %2.3f\tRegLoss %2.3f\tAccuracy %2.3f\tLr %2.5f\tCl %2.3f' % + # (epoch, batch_number + 1, args.epoch_size, duration, loss_, circle_loss_mean_, np.sum(reg_losses_), + # accuracy_, lr_, center_loss_)) + + batch_number += 1 + train_time += duration + summary = tf.Summary() + summary.value.add(tag='time/total', simple_value=train_time) + summary_writer.add_summary(summary, global_step=step_) + return True + +def find_threshold(var, percentile): + hist, bin_edges = np.histogram(var, 100) + cdf = np.float32(np.cumsum(hist)) / np.sum(hist) + bin_centers = (bin_edges[:-1] + bin_edges[1:]) / 2 + threshold = np.interp(percentile * 0.01, cdf, bin_centers) + return threshold + + +def filter_dataset(dataset, data_filename, percentile, min_nrof_images_per_class): + with h5py.File(data_filename, 'r') as f: + distance_to_center = np.array(f.get('distance_to_center')) + label_list = np.array(f.get('label_list')) + image_list = np.array(f.get('image_list')) + distance_to_center_threshold = find_threshold(distance_to_center, percentile) + indices = np.where(distance_to_center >= distance_to_center_threshold)[0] + filtered_dataset = dataset + removelist = [] + for i in indices: + label = label_list[i] + image = image_list[i] + if image in filtered_dataset[label].image_paths: + filtered_dataset[label].image_paths.remove(image) + if len(filtered_dataset[label].image_paths) < min_nrof_images_per_class: + removelist.append(label) + + ix = sorted(list(set(removelist)), reverse=True) + for i in ix: + del (filtered_dataset[i]) + + return filtered_dataset + +def evaluate(sess,iteration,lfw_paths,image_paths_placeholder,labels_placeholder, + phase_train_placeholder, + embeddings,label_batch, actual_issame, batch_size, nrof_folds, log_dir, step, summary_writer,stat, epoch, + distance_metric, subtract_mean, use_flipped_images, + use_fixed_image_standardization): + start_time = time.time() + print('Runnning forward pass on LFW images') + + nrof_embeddings = len(actual_issame) * 2 + nrof_images = nrof_embeddings + + embedding_size = int(embeddings.get_shape()[1]) + + assert nrof_images % batch_size == 0, 'The number of LFW images must be an integer multiple of the LFW batch size' + nrof_batches = nrof_images // batch_size + emb_array = np.zeros((nrof_images, embedding_size)) + lab_array = np.zeros((nrof_images,)) + feed_dict={image_paths_placeholder:lfw_paths, + labels_placeholder:np.array(range(len(lfw_paths)))} + sess.run(iteration.initializer,feed_dict=feed_dict) + + for i in range(nrof_batches): + feed_dict = {phase_train_placeholder: False} + emb,lab= sess.run([embeddings,label_batch], feed_dict=feed_dict) + + lab_array[lab] = lab + emb_array[lab, :] = emb + if i % 10 == 9: + print('.', end='') + sys.stdout.flush() + print('') + embeddings = np.zeros((nrof_embeddings, embedding_size)) + if use_flipped_images: + embeddings[:, :embedding_size] = emb_array[0::2, :] + embeddings[:, embedding_size:] = emb_array[1::2, :] + else: + embeddings = emb_array + + assert np.array_equal(lab_array, np.arange( + nrof_images)) == True, 'Wrong labels used for evaluation, possibly caused by training examples left in the input pipeline' + _, _, accuracy, val, val_std, far = lfw.evaluate(embeddings, actual_issame, nrof_folds=nrof_folds, + distance_metric=distance_metric, subtract_mean=subtract_mean) + fout = open("result.txt", "a+") + print('Accuracy: %2.5f+-%2.5f' % (np.mean(accuracy), np.std(accuracy))) + print('Validation rate: %2.5f+-%2.5f @ FAR=%2.5f' % (val, val_std, far)) + lfw_time = time.time() - start_time + summary = tf.Summary() + summary.value.add(tag='lfw/accuracy', simple_value=np.mean(accuracy)) + summary.value.add(tag='lfw/val_rate', simple_value=val) + summary.value.add(tag='time/lfw', simple_value=lfw_time) + summary_writer.add_summary(summary, step) + with open(os.path.join(log_dir, 'lfw_result.txt'), 'at') as f: + f.write('%d\t%.5f\t%.5f\n' % (step, np.mean(accuracy), val)) + stat['lfw_accuracy'][epoch - 1] = np.mean(accuracy) + stat['lfw_valrate'][epoch - 1] = val + return (np.mean(accuracy), np.std(accuracy)) + +def save_variables_and_metagraph(sess, saver, summary_writer, model_dir, model_name, step): + # print('Saving variables') + start_time = time.time() + checkpoint_path = os.path.join(model_dir, 'model-%s.ckpt' % model_name) + saver.save(sess, checkpoint_path, global_step=step, write_meta_graph=False) + save_time_variables = time.time() - start_time + # print('Variables saved in %.2f seconds' % save_time_variables) + metagraph_filename = os.path.join(model_dir, 'model-%s.meta' % model_name) + save_time_metagraph = 0 + if not os.path.exists(metagraph_filename): + # print('Saving metagraph') + start_time = time.time() + saver.export_meta_graph(metagraph_filename) + save_time_metagraph = time.time() - start_time + # print('Metagraph saved in %.2f seconds' % save_time_metagraph) + summary = tf.Summary() + summary.value.add(tag='time/save_variables', simple_value=save_time_variables) + summary.value.add(tag='time/save_metagraph', simple_value=save_time_metagraph) + summary_writer.add_summary(summary, step) + + +def parse_arguments(argv): + parser = argparse.ArgumentParser() + parser.add_argument('--logs_base_dir', type=str, + help='Directory where to write event logs.', default='.\\logs') + parser.add_argument('--models_base_dir', type=str, + help='Directory where to write trained models and checkpoints.', default='.\\models') + parser.add_argument('--gpu_memory_fraction', type=float, + help='Upper bound on the amount of GPU memory that will be used by the process.', default=0.9) + parser.add_argument('--pretrained_model', type=str, + help='Load a pretrained model before training starts.',default=".\\20220602-013518\\model-20220602-013518.ckpt-258") + parser.add_argument('--data_dir', type=str, + help='Path to the data directory containing aligned face patches.', + default='D:\\pycode\\dataset\\MS-Celeb-1M_clean_align') + parser.add_argument('--model_def', type=str, + help='Model definition. Points to a module containing the definition of the inference graph.', + default='models.resnet34') + parser.add_argument('--max_nrof_epochs', type=int, + help='Number of epochs to run.', default=1) + parser.add_argument('--batch_size', type=int, + help='Number of images to process in a batch.', default=30) + parser.add_argument('--image_size', type=int, + help='Image size (height, width) in pixels.', default=112) + parser.add_argument('--epoch_size', type=int, + help='Number of batches per epoch.', default=1000) + parser.add_argument('--embedding_size', type=int, + help='Dimensionality of the embedding.', default=512) + parser.add_argument('--random_crop', + help='Performs random cropping of training images. If false, the center image_size pixels from the training images are used. ' + + 'If the size of the images in the data directory is equal to image_size no cropping is performed', + action='store_true') + parser.add_argument('--random_flip', + help='Performs random horizontal flipping of training images.', action='store_true') + parser.add_argument('--random_rotate', + help='Performs random rotations of training images.', action='store_true') + parser.add_argument('--use_fixed_image_standardization', + help='Performs fixed standardization of images.', action='store_true') + parser.add_argument('--keep_probability', type=float, + help='Keep probability of dropout for the fully connected layer(s).', default=1.0) + parser.add_argument('--weight_decay', type=float, + help='L2 weight regularization.', default=0.0) + parser.add_argument('--center_loss_factor', type=float, + help='Center loss factor.', default=0.0) + parser.add_argument('--center_loss_alfa', type=float, + help='Center update rate for center loss.', default=0.95) + parser.add_argument('--prelogits_norm_loss_factor', type=float, + help='Loss based on the norm of the activations in the prelogits layer.', default=0.0) + parser.add_argument('--prelogits_norm_p', type=float, + help='Norm to use for prelogits norm loss.', default=1.0) + parser.add_argument('--prelogits_hist_max', type=float, + help='The max value for the prelogits histogram.', default=10.0) + parser.add_argument('--optimizer', type=str, choices=['ADAGRAD', 'ADADELTA', 'ADAM', 'RMSPROP', 'MOM'], + help='The optimization algorithm to use', default='ADAM') + parser.add_argument('--learning_rate', type=float, + help='Initial learning rate. If set to a negative value a learning rate ' + + 'schedule can be specified in the file "learning_rate_schedule.txt"', default=-1.0) + parser.add_argument('--learning_rate_decay_epochs', type=int, + help='Number of epochs between learning rate decay.', default=300) + parser.add_argument('--learning_rate_decay_factor', type=float, + help='Learning rate decay factor.', default=0.1) + parser.add_argument('--moving_average_decay', type=float, + help='Exponential decay for tracking of training parameters.', default=0.9999) + parser.add_argument('--seed', type=int, + help='Random seed.', default=666) + parser.add_argument('--nrof_preprocess_threads', type=int, + help='Number of preprocessing (data loading and augmentation) threads.', default=4) + parser.add_argument('--log_histograms', + help='Enables logging of weight/bias histograms in tensorboard.', action='store_true') + parser.add_argument('--learning_rate_schedule_file', type=str, + help='File containing the learning rate schedule that is used when learning_rate is set to to -1.', + default='.\\data\\learning_rate_circle_loss.txt') + parser.add_argument('--filter_filename', type=str, + help='File containing image data used for dataset filtering', default='') + parser.add_argument('--filter_percentile', type=float, + help='Keep only the percentile images closed to its class center', default=100.0) + parser.add_argument('--filter_min_nrof_images_per_class', type=int, + help='Keep only the classes with this number of examples or more', default=0) + parser.add_argument('--validate_every_n_epochs', type=int, + help='Number of epoch between validation', default=5) + parser.add_argument('--validation_set_split_ratio', type=float, + help='The ratio of the total dataset to use for validation', default=0.0) + parser.add_argument('--min_nrof_val_images_per_class', type=float, + help='Classes with fewer images will be removed from the validation set', default=0) + + parser.add_argument('--lfw_pairs', type=str, + help='The file containing the pairs to use for validation.', default='.\\data\\pairs.txt') + parser.add_argument('--lfw_dir', type=str, + help='Path to the data directory containing aligned face patches.', default='E:\\dataset\\lfw-deepfunneled_align') + parser.add_argument('--lfw_batch_size', type=int, + help='Number of images to process in a batch in the LFW test set.', default=30) + parser.add_argument('--lfw_nrof_folds', type=int, + help='Number of folds to use for cross validation. Mainly used for testing.', default=10) + parser.add_argument('--lfw_distance_metric', type=int, + help='Type of distance metric to use. 0: Euclidian, 1:Cosine similarity distance.', default=0) + parser.add_argument('--lfw_use_flipped_images', + help='Concatenates embeddings for the image and its horizontally flipped counterpart.', + action='store_true') + parser.add_argument('--lfw_subtract_mean', + help='Subtract feature mean before calculating distance.', action='store_true') + return parser.parse_args(argv) + +if __name__ == '__main__': + main(parse_arguments(sys.argv[1:])) diff --git a/TensorFlow/contrib/cv/CircleLoss_ID1221_for_TensorFlow/test_npu.py b/TensorFlow/contrib/cv/CircleLoss_ID1221_for_TensorFlow/test_npu.py new file mode 100644 index 0000000000000000000000000000000000000000..f434095759ab201ab4a8630ce53482a4120f87d6 --- /dev/null +++ b/TensorFlow/contrib/cv/CircleLoss_ID1221_for_TensorFlow/test_npu.py @@ -0,0 +1,523 @@ +# Copyright 2017 The TensorFlow Authors. All Rights Reserved. +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. +# ============================================================================ +# Copyright 2021 Huawei Technologies Co., Ltd +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. +from __future__ import absolute_import +from __future__ import division +from __future__ import print_function +from npu_bridge.npu_init import * + +from datetime import datetime +import os.path +import time +import sys +import random +import tensorflow.compat.v1 as tf +tf.disable_eager_execution() +import numpy as np +import importlib +import argparse +import facenet +import lfw +import h5py +import math +from circle_loss import * +from facenet import load_data +from PIL import Image + +def fully_connected(x, units, activation_fn=None,reuse=False): + out=tf.layers.dense(inputs=x,units=units,activation=activation_fn,reuse=reuse) + return out + +def load_and_preprocess_image(img_path): + img_raw = tf.io.read_file(img_path) + img_tensor = tf.cond( + tf.image.is_jpeg(img_raw), + lambda: tf.image.decode_jpeg(img_raw,3), + lambda: tf.image.decode_png(img_raw,3)) + img_tensor = tf.image.resize(img_tensor, [112, 112]) + img_tensor = tf.cast(img_tensor, tf.float32) + img_tensor = (img_tensor - 127.5) / 128.0 + return img_tensor + +def getIteration(args, image_paths_placeholder, labels_placeholder): + image_paths_dataset=tf.data.Dataset.from_tensor_slices(image_paths_placeholder).map(load_and_preprocess_image) + label_dataset=tf.data.Dataset.from_tensor_slices(labels_placeholder) + dataset=tf.data.Dataset.zip((image_paths_dataset,label_dataset)).prefetch( + args.batch_size).batch(args.batch_size, drop_remainder=True) + iteration=dataset.make_initializable_iterator() + return iteration + +def main(args): + network = importlib.import_module(args.model_def) + subdir = datetime.strftime(datetime.now(), '%Y%m%d-%H%M%S') + log_dir = os.path.join(os.path.expanduser(args.logs_base_dir), subdir) + + if not os.path.isdir(log_dir): + os.makedirs(log_dir) + model_dir = os.path.join(os.path.expanduser(args.models_base_dir), subdir) + if not os.path.isdir(model_dir): + os.makedirs(model_dir) + + stat_file_name = os.path.join(log_dir, 'stat.h5') + + facenet.write_arguments_to_file(args, os.path.join(log_dir, 'arguments.txt')) + + src_path, _ = os.path.split(os.path.realpath(__file__)) + facenet.store_revision_info(src_path, log_dir, ' '.join(sys.argv)) + + np.random.seed(seed=args.seed) + random.seed(args.seed) + dataset = facenet.get_dataset(args.data_dir) + if args.filter_filename: + dataset = filter_dataset(dataset, os.path.expanduser(args.filter_filename), + args.filter_percentile, args.filter_min_nrof_images_per_class) + + if args.validation_set_split_ratio > 0.0: + train_set, val_set = facenet.split_dataset(dataset, args.validation_set_split_ratio, + args.min_nrof_val_images_per_class, 'SPLIT_IMAGES') + else: + train_set, val_set = dataset, [] + + nrof_classes = len(train_set) + + print('Model directory: %s' % model_dir) + #print('Log directory: %s' % log_dir) + pretrained_model = None + if args.pretrained_model: + pretrained_model = os.path.expanduser(args.pretrained_model) + print('Pre-trained model: %s' % pretrained_model) + + if args.lfw_dir: + print('LFW directory: %s' % args.lfw_dir) + pairs = lfw.read_pairs(os.path.expanduser(args.lfw_pairs)) + lfw_paths, actual_issame = lfw.get_paths(os.path.expanduser(args.lfw_dir), pairs) + + with tf.Graph().as_default(): + tf.set_random_seed(args.seed) + global_step = tf.Variable(0, trainable=False) + + image_list, label_list = facenet.get_image_paths_and_labels(train_set) + assert len(image_list) > 0, 'The training set should not be empty' + + val_image_list, val_label_list = facenet.get_image_paths_and_labels(val_set) + + learning_rate_placeholder = tf.placeholder(tf.float32, name='learning_rate') + phase_train_placeholder = tf.placeholder(tf.bool, name='phase_train') + image_paths_placeholder=tf.placeholder(tf.string,shape=(None,),name="image_paths") + labels_placeholder=tf.placeholder(tf.int32,shape=(None,),name="labels") + + iteration = getIteration(args, image_paths_placeholder, labels_placeholder) + image_batch,label_batch=iteration.get_next() + + image_batch = tf.identity(image_batch, 'image_batch') + image_batch = tf.identity(image_batch, 'input') + label_batch = tf.identity(label_batch, 'label_batch') + + # print('Number of classes in training set: %d' % nrof_classes) + # print('Number of examples in training set: %d' % len(image_list)) + # + # print('Number of classes in validation set: %d' % len(val_set)) + # print('Number of examples in validation set: %d' % len(val_image_list)) + + # print('Building training graph') + + prelogits, _ = network.inference(image_batch, args.keep_probability, + phase_train=phase_train_placeholder, bottleneck_layer_size=args.embedding_size, + weight_decay=args.weight_decay) + logits=fully_connected(prelogits,len(train_set),None,False) + + norm_feat = tf.nn.l2_normalize(logits, 1, 1e-10, name='norm_feat') + + circle_loss_mean = circle_loss(tf.one_hot(label_batch, len(train_set), dtype=tf.float32),norm_feat,batch_size=args.batch_size) + + embeddings = tf.nn.l2_normalize(prelogits, 1, 1e-10, name='embeddings') + + + eps = 1e-4 + prelogits_norm = tf.reduce_mean(tf.norm(tf.abs(prelogits) + eps, ord=args.prelogits_norm_p, axis=1)) + tf.add_to_collection(tf.GraphKeys.REGULARIZATION_LOSSES, prelogits_norm * args.prelogits_norm_loss_factor) + + prelogits_center_loss, _ = facenet.center_loss(prelogits, label_batch, args.center_loss_alfa, nrof_classes) + tf.add_to_collection(tf.GraphKeys.REGULARIZATION_LOSSES, prelogits_center_loss * args.center_loss_factor) + + learning_rate = tf.train.exponential_decay(learning_rate_placeholder, global_step, + args.learning_rate_decay_epochs * args.epoch_size, + args.learning_rate_decay_factor, staircase=True) + tf.summary.scalar('learning_rate', learning_rate) + + tf.add_to_collection('losses', circle_loss_mean) + + correct_prediction = tf.cast(tf.equal(tf.argmax(logits, 1), tf.cast(label_batch, tf.int64)), tf.float32) + accuracy = tf.reduce_mean(correct_prediction) + + regularization_losses = tf.get_collection(tf.GraphKeys.REGULARIZATION_LOSSES) + + total_loss = tf.add_n([circle_loss_mean] + regularization_losses, name='total_loss') + + train_op = facenet.train(total_loss, global_step, args.optimizer, + learning_rate, args.moving_average_decay, tf.global_variables(), args.log_histograms) + + saver = tf.train.Saver(tf.trainable_variables(), max_to_keep=3) + + summary_op = tf.summary.merge_all() + + gpu_options = tf.GPUOptions(per_process_gpu_memory_fraction=args.gpu_memory_fraction) + + sess = tf.Session(config=npu_config_proto(config_proto=tf.ConfigProto(gpu_options=gpu_options, log_device_placement=False))) + sess.run(tf.global_variables_initializer()) + sess.run(tf.local_variables_initializer()) + + summary_writer = tf.summary.FileWriter(log_dir, sess.graph) + coord = tf.train.Coordinator() + tf.train.start_queue_runners(coord=coord, sess=sess) + + with sess.as_default(): + + if pretrained_model: + print('Restoring pretrained model: %s' % pretrained_model) + saver.restore(sess, pretrained_model) + + # print('Running training') + nrof_steps = args.max_nrof_epochs * args.epoch_size + nrof_val_samples = int(math.ceil( + args.max_nrof_epochs / args.validate_every_n_epochs)) + stat = { + 'loss': np.zeros((nrof_steps,), np.float32), + 'center_loss': np.zeros((nrof_steps,), np.float32), + 'reg_loss': np.zeros((nrof_steps,), np.float32), + 'xent_loss': np.zeros((nrof_steps,), np.float32), + 'prelogits_norm': np.zeros((nrof_steps,), np.float32), + 'accuracy': np.zeros((nrof_steps,), np.float32), + 'val_loss': np.zeros((nrof_val_samples,), np.float32), + 'val_xent_loss': np.zeros((nrof_val_samples,), np.float32), + 'val_accuracy': np.zeros((nrof_val_samples,), np.float32), + 'lfw_accuracy': np.zeros((args.max_nrof_epochs,), np.float32), + 'lfw_valrate': np.zeros((args.max_nrof_epochs,), np.float32), + 'learning_rate': np.zeros((args.max_nrof_epochs,), np.float32), + 'time_train': np.zeros((args.max_nrof_epochs,), np.float32), + 'time_validate': np.zeros((args.max_nrof_epochs,), np.float32), + 'time_evaluate': np.zeros((args.max_nrof_epochs,), np.float32), + 'prelogits_hist': np.zeros((args.max_nrof_epochs, 1000), np.float32), + } + + for epoch in range(1, args.max_nrof_epochs + 1): + step = sess.run(global_step, feed_dict=None) + + save_variables_and_metagraph(sess, saver, summary_writer, model_dir, subdir, epoch) + + t = time.time() + if args.lfw_dir: + acc_mean,acc_var=evaluate(sess,iteration,lfw_paths, image_paths_placeholder, labels_placeholder, phase_train_placeholder, + embeddings,label_batch,actual_issame, args.lfw_batch_size, + args.lfw_nrof_folds, log_dir, step, summary_writer,stat,epoch, + args.lfw_distance_metric, args.lfw_subtract_mean, args.lfw_use_flipped_images, + args.use_fixed_image_standardization) + fout=open("result.txt","a+") + fout.write(str(epoch)+","+str(acc_mean)+"+-"+str(acc_var)+"\n") + fout.close() + stat['time_evaluate'][epoch - 1] = time.time() - t + + # print('Saving statistics') + with h5py.File(stat_file_name, 'w') as f: + for key, value in stat.items(): + f.create_dataset(key, data=value) + + return model_dir + +def train_epoch(args,sess,epoch,iteration,image_list,label_list,image_paths_placeholder,labels_placeholder, + learning_rate_placeholder,phase_train_placeholder,step, + loss,train_op,summary_op,summary_writer,reg_losses,learning_rate_schedule_file, + stat,circle_loss_mean,accuracy, + learning_rate,prelogits,prelogits_center_loss,random_rotate,random_crop,random_flip,prelogits_norm, + prelogits_hist_max,use_fixed_image_standardization): + + epoch_idx=np.array(random.sample(range(len(label_list)),args.epoch_size*args.batch_size)) + image_list=np.array(image_list)[epoch_idx] + label_list=np.array(label_list)[epoch_idx] + + batch_number = 0 + if args.learning_rate > 0.0: + lr = args.learning_rate + else: + lr = facenet.get_learning_rate_from_file(learning_rate_schedule_file, epoch) + + if lr <= 0: + return False + feed_dict={image_paths_placeholder:image_list, + labels_placeholder:label_list} + sess.run(iteration.initializer,feed_dict=feed_dict) + + train_time = 0 + while batch_number < args.epoch_size: + start_time = time.time() + + feed_dict = {learning_rate_placeholder: lr, + phase_train_placeholder: True} + + tensor_list = [loss, train_op, step, reg_losses, prelogits, circle_loss_mean, learning_rate, prelogits_norm, + accuracy, prelogits_center_loss] + if batch_number % 100 == 0: + loss_, _, step_, reg_losses_, prelogits_, circle_loss_mean_, lr_, prelogits_norm_, accuracy_, center_loss_, summary_str = sess.run(tensor_list + [summary_op], feed_dict=feed_dict) + summary_writer.add_summary(summary_str, global_step=step_) + else: + loss_, _, step_, reg_losses_, prelogits_, circle_loss_mean_, lr_, prelogits_norm_, accuracy_, center_loss_ = sess.run(tensor_list, feed_dict=feed_dict) + + stat['loss'][step_ - 1] = loss_ + stat['center_loss'][step_ - 1] = center_loss_ + stat['reg_loss'][step_ - 1] = np.sum(reg_losses_) + stat['xent_loss'][step_ - 1] = circle_loss_mean_ + stat['prelogits_norm'][step_ - 1] = prelogits_norm_ + stat['learning_rate'][epoch - 1] = lr_ + stat['accuracy'][step_ - 1] = accuracy_ + stat['prelogits_hist'][epoch - 1, :] += \ + np.histogram(np.minimum(np.abs(prelogits_), prelogits_hist_max), bins=1000, range=(0.0, prelogits_hist_max))[0] + + duration = time.time() - start_time + # print( + # 'Epoch: [%d][%d/%d]\tTime %.3f\tLoss %2.3f\tXent %2.3f\tRegLoss %2.3f\tAccuracy %2.3f\tLr %2.5f\tCl %2.3f' % + # (epoch, batch_number + 1, args.epoch_size, duration, loss_, circle_loss_mean_, np.sum(reg_losses_), + # accuracy_, lr_, center_loss_)) + + batch_number += 1 + train_time += duration + summary = tf.Summary() + summary.value.add(tag='time/total', simple_value=train_time) + summary_writer.add_summary(summary, global_step=step_) + return True + +def find_threshold(var, percentile): + hist, bin_edges = np.histogram(var, 100) + cdf = np.float32(np.cumsum(hist)) / np.sum(hist) + bin_centers = (bin_edges[:-1] + bin_edges[1:]) / 2 + threshold = np.interp(percentile * 0.01, cdf, bin_centers) + return threshold + + +def filter_dataset(dataset, data_filename, percentile, min_nrof_images_per_class): + with h5py.File(data_filename, 'r') as f: + distance_to_center = np.array(f.get('distance_to_center')) + label_list = np.array(f.get('label_list')) + image_list = np.array(f.get('image_list')) + distance_to_center_threshold = find_threshold(distance_to_center, percentile) + indices = np.where(distance_to_center >= distance_to_center_threshold)[0] + filtered_dataset = dataset + removelist = [] + for i in indices: + label = label_list[i] + image = image_list[i] + if image in filtered_dataset[label].image_paths: + filtered_dataset[label].image_paths.remove(image) + if len(filtered_dataset[label].image_paths) < min_nrof_images_per_class: + removelist.append(label) + + ix = sorted(list(set(removelist)), reverse=True) + for i in ix: + del (filtered_dataset[i]) + + return filtered_dataset + +def evaluate(sess,iteration,lfw_paths,image_paths_placeholder,labels_placeholder, + phase_train_placeholder, + embeddings,label_batch, actual_issame, batch_size, nrof_folds, log_dir, step, summary_writer,stat, epoch, + distance_metric, subtract_mean, use_flipped_images, + use_fixed_image_standardization): + start_time = time.time() + print('Runnning forward pass on LFW images') + + nrof_embeddings = len(actual_issame) * 2 + nrof_images = nrof_embeddings + + embedding_size = int(embeddings.get_shape()[1]) + + assert nrof_images % batch_size == 0, 'The number of LFW images must be an integer multiple of the LFW batch size' + nrof_batches = nrof_images // batch_size + emb_array = np.zeros((nrof_images, embedding_size)) + lab_array = np.zeros((nrof_images,)) + feed_dict={image_paths_placeholder:lfw_paths, + labels_placeholder:np.array(range(len(lfw_paths)))} + sess.run(iteration.initializer,feed_dict=feed_dict) + + for i in range(nrof_batches): + feed_dict = {phase_train_placeholder: False} + emb,lab= sess.run([embeddings,label_batch], feed_dict=feed_dict) + + lab_array[lab] = lab + emb_array[lab, :] = emb + if i % 10 == 9: + print('.', end='') + sys.stdout.flush() + print('') + embeddings = np.zeros((nrof_embeddings, embedding_size)) + if use_flipped_images: + embeddings[:, :embedding_size] = emb_array[0::2, :] + embeddings[:, embedding_size:] = emb_array[1::2, :] + else: + embeddings = emb_array + + assert np.array_equal(lab_array, np.arange( + nrof_images)) == True, 'Wrong labels used for evaluation, possibly caused by training examples left in the input pipeline' + _, _, accuracy, val, val_std, far = lfw.evaluate(embeddings, actual_issame, nrof_folds=nrof_folds, + distance_metric=distance_metric, subtract_mean=subtract_mean) + fout = open("result.txt", "a+") + print('Accuracy: %2.5f+-%2.5f' % (np.mean(accuracy), np.std(accuracy))) + print('Validation rate: %2.5f+-%2.5f @ FAR=%2.5f' % (val, val_std, far)) + lfw_time = time.time() - start_time + summary = tf.Summary() + summary.value.add(tag='lfw/accuracy', simple_value=np.mean(accuracy)) + summary.value.add(tag='lfw/val_rate', simple_value=val) + summary.value.add(tag='time/lfw', simple_value=lfw_time) + summary_writer.add_summary(summary, step) + with open(os.path.join(log_dir, 'lfw_result.txt'), 'at') as f: + f.write('%d\t%.5f\t%.5f\n' % (step, np.mean(accuracy), val)) + stat['lfw_accuracy'][epoch - 1] = np.mean(accuracy) + stat['lfw_valrate'][epoch - 1] = val + return (np.mean(accuracy), np.std(accuracy)) + +def save_variables_and_metagraph(sess, saver, summary_writer, model_dir, model_name, step): + # print('Saving variables') + start_time = time.time() + checkpoint_path = os.path.join(model_dir, 'model-%s.ckpt' % model_name) + saver.save(sess, checkpoint_path, global_step=step, write_meta_graph=False) + save_time_variables = time.time() - start_time + # print('Variables saved in %.2f seconds' % save_time_variables) + metagraph_filename = os.path.join(model_dir, 'model-%s.meta' % model_name) + save_time_metagraph = 0 + if not os.path.exists(metagraph_filename): + # print('Saving metagraph') + start_time = time.time() + saver.export_meta_graph(metagraph_filename) + save_time_metagraph = time.time() - start_time + # print('Metagraph saved in %.2f seconds' % save_time_metagraph) + summary = tf.Summary() + summary.value.add(tag='time/save_variables', simple_value=save_time_variables) + summary.value.add(tag='time/save_metagraph', simple_value=save_time_metagraph) + summary_writer.add_summary(summary, step) + + +def parse_arguments(argv): + parser = argparse.ArgumentParser() + parser.add_argument('--logs_base_dir', type=str, + help='Directory where to write event logs.', default='./logs') + parser.add_argument('--models_base_dir', type=str, + help='Directory where to write trained models and checkpoints.', default='./models') + parser.add_argument('--gpu_memory_fraction', type=float, + help='Upper bound on the amount of GPU memory that will be used by the process.', default=0.9) + parser.add_argument('--pretrained_model', type=str, + help='Load a pretrained model before training starts.',default="./20220602-013518/model-20220602-013518.ckpt-258") + parser.add_argument('--data_dir', type=str, + help='Path to the data directory containing aligned face patches.', + default='./dataset/MS-Celeb-1M_clean_align') + parser.add_argument('--model_def', type=str, + help='Model definition. Points to a module containing the definition of the inference graph.', + default='models.resnet34') + parser.add_argument('--max_nrof_epochs', type=int, + help='Number of epochs to run.', default=1) + parser.add_argument('--batch_size', type=int, + help='Number of images to process in a batch.', default=30) + parser.add_argument('--image_size', type=int, + help='Image size (height, width) in pixels.', default=112) + parser.add_argument('--epoch_size', type=int, + help='Number of batches per epoch.', default=1000) + parser.add_argument('--embedding_size', type=int, + help='Dimensionality of the embedding.', default=512) + parser.add_argument('--random_crop', + help='Performs random cropping of training images. If false, the center image_size pixels from the training images are used. ' + + 'If the size of the images in the data directory is equal to image_size no cropping is performed', + action='store_true') + parser.add_argument('--random_flip', + help='Performs random horizontal flipping of training images.', action='store_true') + parser.add_argument('--random_rotate', + help='Performs random rotations of training images.', action='store_true') + parser.add_argument('--use_fixed_image_standardization', + help='Performs fixed standardization of images.', action='store_true') + parser.add_argument('--keep_probability', type=float, + help='Keep probability of dropout for the fully connected layer(s).', default=1.0) + parser.add_argument('--weight_decay', type=float, + help='L2 weight regularization.', default=0.0) + parser.add_argument('--center_loss_factor', type=float, + help='Center loss factor.', default=0.0) + parser.add_argument('--center_loss_alfa', type=float, + help='Center update rate for center loss.', default=0.95) + parser.add_argument('--prelogits_norm_loss_factor', type=float, + help='Loss based on the norm of the activations in the prelogits layer.', default=0.0) + parser.add_argument('--prelogits_norm_p', type=float, + help='Norm to use for prelogits norm loss.', default=1.0) + parser.add_argument('--prelogits_hist_max', type=float, + help='The max value for the prelogits histogram.', default=10.0) + parser.add_argument('--optimizer', type=str, choices=['ADAGRAD', 'ADADELTA', 'ADAM', 'RMSPROP', 'MOM'], + help='The optimization algorithm to use', default='ADAM') + parser.add_argument('--learning_rate', type=float, + help='Initial learning rate. If set to a negative value a learning rate ' + + 'schedule can be specified in the file "learning_rate_schedule.txt"', default=0.01) + parser.add_argument('--learning_rate_decay_epochs', type=int, + help='Number of epochs between learning rate decay.', default=300) + parser.add_argument('--learning_rate_decay_factor', type=float, + help='Learning rate decay factor.', default=0.1) + parser.add_argument('--moving_average_decay', type=float, + help='Exponential decay for tracking of training parameters.', default=0.9999) + parser.add_argument('--seed', type=int, + help='Random seed.', default=666) + parser.add_argument('--nrof_preprocess_threads', type=int, + help='Number of preprocessing (data loading and augmentation) threads.', default=4) + parser.add_argument('--log_histograms', + help='Enables logging of weight/bias histograms in tensorboard.', action='store_true') + parser.add_argument('--learning_rate_schedule_file', type=str, + help='File containing the learning rate schedule that is used when learning_rate is set to to -1.', + default='./data/learning_rate_circle_loss.txt') + parser.add_argument('--filter_filename', type=str, + help='File containing image data used for dataset filtering', default='') + parser.add_argument('--filter_percentile', type=float, + help='Keep only the percentile images closed to its class center', default=100.0) + parser.add_argument('--filter_min_nrof_images_per_class', type=int, + help='Keep only the classes with this number of examples or more', default=0) + parser.add_argument('--validate_every_n_epochs', type=int, + help='Number of epoch between validation', default=5) + parser.add_argument('--validation_set_split_ratio', type=float, + help='The ratio of the total dataset to use for validation', default=0.0) + parser.add_argument('--min_nrof_val_images_per_class', type=float, + help='Classes with fewer images will be removed from the validation set', default=0) + + parser.add_argument('--lfw_pairs', type=str, + help='The file containing the pairs to use for validation.', default='./data/pairs.txt') + parser.add_argument('--lfw_dir', type=str, + help='Path to the data directory containing aligned face patches.', default='./dataset/lfw-deepfunneled_align') + parser.add_argument('--lfw_batch_size', type=int, + help='Number of images to process in a batch in the LFW test set.', default=30) + parser.add_argument('--lfw_nrof_folds', type=int, + help='Number of folds to use for cross validation. Mainly used for testing.', default=10) + parser.add_argument('--lfw_distance_metric', type=int, + help='Type of distance metric to use. 0: Euclidian, 1:Cosine similarity distance.', default=0) + parser.add_argument('--lfw_use_flipped_images', + help='Concatenates embeddings for the image and its horizontally flipped counterpart.', + action='store_true') + parser.add_argument('--lfw_subtract_mean', + help='Subtract feature mean before calculating distance.', action='store_true') + return parser.parse_args(argv) + +if __name__ == '__main__': + main(parse_arguments(sys.argv[1:])) + diff --git a/TensorFlow/contrib/cv/CircleLoss_ID1221_for_TensorFlow/train.py b/TensorFlow/contrib/cv/CircleLoss_ID1221_for_TensorFlow/train.py deleted file mode 100644 index bea503f46027e954f3f161cc3a9a5acda47968f7..0000000000000000000000000000000000000000 --- a/TensorFlow/contrib/cv/CircleLoss_ID1221_for_TensorFlow/train.py +++ /dev/null @@ -1,320 +0,0 @@ -# Copyright 2017 The TensorFlow Authors. All Rights Reserved. -# -# Licensed under the Apache License, Version 2.0 (the "License"); -# you may not use this file except in compliance with the License. -# You may obtain a copy of the License at -# -# http://www.apache.org/licenses/LICENSE-2.0 -# -# Unless required by applicable law or agreed to in writing, software -# distributed under the License is distributed on an "AS IS" BASIS, -# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. -# See the License for the specific language governing permissions and -# limitations under the License. -# ============================================================================ -# Copyright 2021 Huawei Technologies Co., Ltd -# -# Licensed under the Apache License, Version 2.0 (the "License"); -# you may not use this file except in compliance with the License. -# You may obtain a copy of the License at -# -# http://www.apache.org/licenses/LICENSE-2.0 -# -# Unless required by applicable law or agreed to in writing, software -# distributed under the License is distributed on an "AS IS" BASIS, -# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. -# See the License for the specific language governing permissions and -# limitations under the License. -import numpy as np -import skimage.transform -import tensorflow as tf -# from config import * -from model import * -from data import * -from loss import * -from tensorflow.contrib.slim import nets -import time -import argparse - - -arg = argparse.ArgumentParser() -timestamp = int(time.time()) -arg.add_argument('-e','--expriment_name',type=str,default=f'test_{timestamp}') -arg.add_argument('-g','--gpu',type=str,default='1') -arg.add_argument('-w','--warmup',type=int,default=5) -arg.add_argument('-b','--batch_size',type=int,default=32) -arg.add_argument('-epoch','--epoch_size',type=int,default=10) -arg.add_argument('-modelname','--modelname',type=str,default='softmaxcircleloss') -arg.add_argument('-lr','--init_lr',type=float,default=0.08) -arg.add_argument('-lamda','--lamda',type=float,default=1) -arg.add_argument('-drop','--droprate',type=float,default=0.5) -arg.add_argument("--train_url", type=str, default="./output") -arg.add_argument("--data_url", type=str, default="./dataset") -arg.add_argument("--ckpt_url", type=str, default="./ckpt") -arg.add_argument("--erasing_prob", type=float, default=0.5) - -args = arg.parse_args() - -def makedirifnotexist(path): - if not os.path.exists(path): - os.makedirs(path) - -homedir = "/home/nanshen/xutan/yixin/circlesoftmax" -log_dir = f'{homedir}/logs/{args.expriment_name}/' -im_size = [256, 128] -market_dir = 'Data/Market-1501-v15.09.15' -warmup = args.warmup -m=0.25 -gamma=256 -lamda = args.lamda -num_classes=751 -epoch_size = args.epoch_size -batch_size = args.batch_size -embedding_size=128 -resnet_model_path = '/home/nanshen/xutan/yixin/circlesoftmax/checkpoint/resnet_v2_50.ckpt' # Path to the pretrained model -model_save_path = log_dir + args.modelname # Path to the model.ckpt-(num_steps) will be saved -ckptpath = os.path.join(log_dir,'ckpt') -makedirifnotexist(ckptpath) -num_bottleneck = 512 -init_lr = args.init_lr -os.environ["CUDA_VISIBLE_DEVICES"]=args.gpu - -x = tf.placeholder(tf.float32, [None, im_size[0],im_size[1],3],name='inputs') -y_ = tf.placeholder(tf.int32, [None],name='labels') -lr_backbone = tf.placeholder(tf.float32, name='lr_backbone') -lr_logit = tf.placeholder(tf.float32, name='lr_logit') -is_training = tf.placeholder(tf.bool, name='is_training') -keep_rate = tf.placeholder(tf.float32, name='keep_rate') - - -# 读入数据 -imnum, dataset,le = tf_dataset(erasing_prob=args.erasing_prob,batch_size=batch_size,epoch_size=None,path='/home/nanshen/xutan/yixin/market1501/Market-1501-v15.09.15/bounding_box_train') -iterator = dataset.make_one_shot_iterator() -one_element = iterator.get_next() - -# with slim.arg_scope(nets.resnet_v1.resnet_arg_scope()): -# net, endpoints = nets.resnet_v1.resnet_v1_50(x, num_classes=None, -# is_training=is_training,global_pool=True) -# from tf_slim.nets import resnet_v2 -with slim.arg_scope(nets.resnet_v2.resnet_arg_scope()): - net, end_points = nets.resnet_v2.resnet_v2_50(x, num_classes=None, is_training=is_training) - -import math -with tf.variable_scope('Logits'): - # batch_norm_params = {"is_training": is_training, "decay": 0.9, "updates_collections": None} - with slim.arg_scope([slim.conv2d, slim.fully_connected], - weights_regularizer=slim.l2_regularizer(5e-4)): - resnetfeat = tf.squeeze(net, axis=[1, 2]) - fc1 = slim.fully_connected(resnetfeat, num_outputs=num_bottleneck, - activation_fn=None,weights_initializer =tf.random_normal_initializer(mean=0,stddev=math.sqrt(2/num_bottleneck))) - bn1 = slim.batch_norm(fc1, activation_fn=None,is_training=is_training) - drop1 = slim.dropout(bn1, keep_prob=keep_rate) - norm_feat = tf.math.l2_normalize(drop1,axis=1,name='features',epsilon=1e-12) - circle_loss_op,sp,sn,recall,precision,f1_score = PairWiseCircleLoss(norm_feat,y_) - tf.summary.histogram('sn',sn) - tf.summary.histogram('sp',sp) - logits = slim.fully_connected(drop1, num_outputs=num_classes, - activation_fn=None,weights_initializer=tf.random_normal_initializer(mean=0,stddev=0.001)) - -checkpoint_exclude_scopes = 'Logits' -exclusions = None -if checkpoint_exclude_scopes: - exclusions = [ - scope.strip() for scope in checkpoint_exclude_scopes.split(',')] -variables_to_restore = [] -add_var = [] -outof = [] - -for var in slim.get_model_variables(): - excluded = False - for exclusion in exclusions: - if var.op.name.startswith(exclusion): - excluded = True - if not excluded: - variables_to_restore.append(var) - else: - add_var.append(var) -# regularizer = tf.contrib.layers.l2_regularizer(5e-4) -# reg_term = tf.contrib.layers.apply_regularization(regularizer,weights_list=add_var) - -# softmaxloss = tf.nn.sparse_softmax_cross_entropy_with_logits( -# labels=y_, logits=logits) -# softmaxloss = tf.reduce_mean(softmaxloss) -# soft_logit = tf.clip_by_value(soft_logit, 0.000001, 0.999999) -labels = tf.one_hot(y_,num_classes) -soft_logit = tf.nn.softmax(logits) -# softmaxloss = tf.nn.softmax_cross_entropy_with_logits(labels=labels, logits=logits) -soft_logits = tf.clip_by_value(soft_logit, 0.000001, 0.999999) -cross_entropy = -tf.reduce_sum(labels * tf.log(soft_logits), axis=1) -softmaxloss = tf.reduce_mean(cross_entropy) -circle_loss_term = lamda * circle_loss_op/batch_size -loss = softmaxloss + circle_loss_term - - -soft_logit = tf.nn.softmax(logits) -classes = tf.argmax(soft_logit, axis=1, name='classes') -accuracy_softmax = tf.reduce_mean(tf.cast( - tf.equal(tf.cast(classes, dtype=tf.int32), y_), dtype=tf.float32)) - -# optimizer1 = tf.train.AdamOptimizer(learning_rate=lr_backbone) -# optimizer2 = tf.train.AdamOptimizer(learning_rate=lr_logit) -optimizer1 = tf.train.MomentumOptimizer(learning_rate=lr_backbone,momentum=0.9,use_nesterov=True) -optimizer2 = tf.train.MomentumOptimizer(learning_rate=lr_logit,momentum=0.9,use_nesterov=True) - -# bn var -update_ops = tf.get_collection(tf.GraphKeys.UPDATE_OPS) -with tf.control_dependencies(update_ops): - train_step1 = optimizer1.minimize(loss,var_list=variables_to_restore) - train_step2 = optimizer2.minimize(loss,var_list=add_var) - -init = tf.global_variables_initializer() - -saver_restore = tf.train.Saver(var_list=variables_to_restore) -saver = tf.train.Saver(tf.global_variables()) - -config = tf.ConfigProto(allow_soft_placement = True) -config.gpu_options.allow_growth = True #程序按需申请内存 - -# tensorboard记录下来 - -# recall,precision,f1_score -tf.summary.scalar('pairwise_recall',recall) -tf.summary.scalar('pairwise_precision',precision) -tf.summary.scalar('pairwise_f1_score',f1_score) -tf.summary.scalar('softmaxloss',softmaxloss) -# tf.summary.scalar('reg_loss',reg_term) -tf.summary.scalar('circleloss',circle_loss_term) -tf.summary.scalar('totalloss',loss) -tf.summary.scalar('accuracy_softmax',accuracy_softmax) -tf.summary.scalar('lr_backbone',lr_backbone) -tf.summary.scalar('lr_logit',lr_logit) -merged = tf.summary.merge_all() - -def adjust_LR(step,epoch): - lr_l = init_lr * 0.1**((epoch)//40) - lr_b = 0.1 * lr_l - if epoch < warmup: - factor = min(1.0, 0.1 + 0.9 / (imnum//batch_size*warmup)) - lr_l,lr_b = factor*lr_l,factor*lr_b - if step < 32000//args.batch_size: # freeze bonee - lr_b = 0 - return lr_l,lr_b - - -least_graph_path = None -least_param_path = None - - -# 开始训练 -with tf.Session(config=config) as sess: - - # writter = tf.summary.FileWriter(log_dir, sess.graph) - train_writer = tf.summary.FileWriter(log_dir + '/train', sess.graph) - # test_writer = tf.summary.FileWriter(log_dir + '/test',sess.graph) - - # 执行初始化操作 - sess.run(init) - - # Load the pretrained checkpoint file xxx.ckpt - saver_restore.restore(sess, resnet_model_path) - - # for i in range(step_size): - # if i % 500 ==0: - # saver.save(sess, log_dir + "/model.ckpt", i) - # if i % 10 == 0: - # # 每10步测试一下 - # summary = sess.run(merged, feed_dict=feed_dict(marketdata,False,x,y_,lr,lr_new)) - # test_writer.add_summary(summary, i) - # else: - # if i % 100 == 99: - # # 每100次记录训练运算时间和内存占用等信息 - # run_options = tf.RunOptions(trace_level=tf.RunOptions.FULL_TRACE) - # run_metadata = tf.RunMetadata() - # summary, _ = sess.run([merged, train_step], - # feed_dict=feed_dict(marketdata,True,x,y_,lr,lr_new), - # options=run_options, - # run_metadata=run_metadata) - # train_writer.add_run_metadata(run_metadata, 'step%03d' % i) - # train_writer.add_summary(summary, i) - # else: - # # 正常情况只进行汇总和训练步伐操作 - # summary, _ = sess.run([merged, train_step], feed_dict=feed_dict(marketdata,True,x,y_,lr,lr_new)) - # train_writer.add_summary(summary, i) - epoch = 0 - sample_counter = 0 - step_counter = 0 - # for i in range(step_size): - import time - t0=time.time() - while epoch < epoch_size: - step_counter+=1 - # epoch = np.mean(traindata._epochs_completed) - import time - ts = time.time() - images, groundtruth_lists = sess.run(one_element) - # images_test, groundtruth_lists_test = sess.run(testnext) - - # print(groundtruth_lists) - batch_now = images.shape[0] - sample_counter+=batch_now - epoch = sample_counter // imnum - if batch_now < batch_size: # skip the last batch - print('skip last') - continue - lr_l,lr_b = adjust_LR(step_counter,epoch) - - train_dict = {x: images, - y_: groundtruth_lists, - is_training: True, - lr_logit:lr_l, - lr_backbone:lr_b, - keep_rate:0.5} - # test_dict = {x: images_test, - # y_: groundtruth_lists_test, - # is_training: False, - # lr_logit:lr_l, - # lr_backbone:lr_b, - # keep_rate:1.0} - summary,loss_, acc_,f1,softmaxlossterm,circlelossterm,_,_ = sess.run([merged,loss, accuracy_softmax,f1_score, softmaxloss,circle_loss_term,train_step1,train_step2], feed_dict=train_dict) - # summary_test = sess.run(merged, feed_dict=test_dict) - train_writer.add_summary(summary, step_counter) - # test_writer.add_summary(summary_test,step_counter) - - t1 = time.time() - timingperstep = (t1-ts) - timing = int(t1 - t0) - # train_text = 'Epoch:{}, Step: {}, train_Loss: {:.4f}, train_Accuracy: {:.4f}, Timing:{:.4f}s'.format( - # epoch,step_counter+1, loss_, acc_,t1-t0) - print("Epoch:{} Step:{} TrainLoss:{:.4f} SoftmaxLoss:{:.4f} CircleLoss:{:.4f} SoftmaxAcc:{:.4f} PairwiseF1:{:.4f} TimePerStep:{:.4f}s TotalTiming:{}m{}s".format(epoch,step_counter,loss_,softmaxlossterm,circlelossterm,acc_,f1,timingperstep,(timing)//60,timing%60)) - - if (step_counter+1)%(imnum//batch_size) == 0: - ckptpathnew = os.path.join(ckptpath,str(step_counter+1)) - makedirifnotexist(ckptpathnew) - saver.save(sess, os.path.join(ckptpathnew,args.modelname)+'.ckpt', global_step=step_counter+1) - least_graph_path = os.path.join(ckptpathnew,args.modelname)+f'.ckpt-{str(step_counter+1)}.meta' - least_param_path = ckptpathnew - tf.io.write_graph(sess.graph, model_save_path, 'graph.pbtxt', as_text=True) - print('save mode to {}'.format(model_save_path)) - # break - - # if (i+1) % 10 == 0: - # t0=time.time() - # images, groundtruth_lists = get_next_batch(marketdata,train=False) - # test_dict = {x: images, - # y_: groundtruth_lists, - # is_training: True, - # lr:lr_new} - # summary,loss_val,acc_val = sess.run([merged,loss, accuracy], feed_dict=test_dict) - # test_writer.add_summary(summary, i) - # t1 = time.time() - # test_text = 'Step: {}, Loss: {:.4f}, Accuracy: {:.4f}, Timing:{:.4f}s'.format( - # i+1, loss_val, acc_val,t1-t0) - # print(test_text) - -train_writer.close() -print(least_graph_path) -print(least_param_path) -syscall = f'python /home/nanshen/xutan/yixin/circlesoftmax/code/test.py --graphpath {least_graph_path} --parampath {least_param_path} --logdir {log_dir}' -print('测试阶段') -print(syscall) -os.system(syscall) diff --git a/TensorFlow/contrib/cv/CircleLoss_ID1221_for_TensorFlow/train_1p.sh b/TensorFlow/contrib/cv/CircleLoss_ID1221_for_TensorFlow/train_1p.sh new file mode 100644 index 0000000000000000000000000000000000000000..c27956602393fa693423b18c6a8aaf1a90a5210e --- /dev/null +++ b/TensorFlow/contrib/cv/CircleLoss_ID1221_for_TensorFlow/train_1p.sh @@ -0,0 +1,4 @@ +python3 train_npu.py \ + --lfw_dir ./dataset/lfw-deepfunneled_align \ + --data_dir ./dataset/MS-Celeb-1M_clean_align \ + --pretrained_model ./20220602-013518/model-20220602-013518.ckpt-258 \ No newline at end of file diff --git a/TensorFlow/contrib/cv/CircleLoss_ID1221_for_TensorFlow/train_PK.py b/TensorFlow/contrib/cv/CircleLoss_ID1221_for_TensorFlow/train_PK.py deleted file mode 100644 index e7c0122f2286f51a30c3c5441bd8bfcfc23d75e5..0000000000000000000000000000000000000000 --- a/TensorFlow/contrib/cv/CircleLoss_ID1221_for_TensorFlow/train_PK.py +++ /dev/null @@ -1,329 +0,0 @@ -# Copyright 2017 The TensorFlow Authors. All Rights Reserved. -# -# Licensed under the Apache License, Version 2.0 (the "License"); -# you may not use this file except in compliance with the License. -# You may obtain a copy of the License at -# -# http://www.apache.org/licenses/LICENSE-2.0 -# -# Unless required by applicable law or agreed to in writing, software -# distributed under the License is distributed on an "AS IS" BASIS, -# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. -# See the License for the specific language governing permissions and -# limitations under the License. -# ============================================================================ -# Copyright 2021 Huawei Technologies Co., Ltd -# -# Licensed under the Apache License, Version 2.0 (the "License"); -# you may not use this file except in compliance with the License. -# You may obtain a copy of the License at -# -# http://www.apache.org/licenses/LICENSE-2.0 -# -# Unless required by applicable law or agreed to in writing, software -# distributed under the License is distributed on an "AS IS" BASIS, -# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. -# See the License for the specific language governing permissions and -# limitations under the License. -import numpy as np -import skimage.transform -import tensorflow as tf -# from config import * -from model import * -from data import * -from loss import * -from tensorflow.contrib.slim import nets -import time -import argparse - - -arg = argparse.ArgumentParser() -timestamp = int(time.time()) -arg.add_argument('-e','--expriment_name',type=str,default=f'test_{timestamp}') -arg.add_argument('-g','--gpu',type=str,default='0,1') -arg.add_argument('-w','--warmup',type=int,default=5) -# arg.add_argument('-b','--batch_size',type=int,default=32) -arg.add_argument('-epoch','--epoch_size',type=int,default=60) -arg.add_argument('-modelname','--modelname',type=str,default='softmaxcircleloss') -arg.add_argument('-lr','--init_lr',type=float,default=0.02) -arg.add_argument('-lamda','--lamda',type=float,default=1) -arg.add_argument('-drop','--droprate',type=float,default=0.5) -arg.add_argument("--train_url", type=str, default="./output") -arg.add_argument("--data_url", type=str, default="./dataset") -arg.add_argument("--ckpt_url", type=str, default="./ckpt") -arg.add_argument("--erasing_prob", type=float, default=0.5) -arg.add_argument("--P", type=int, default=4) -arg.add_argument("--K", type=int, default=4) - - -args = arg.parse_args() - - -homedir = "/home/nanshen/xutan/yixin/circlesoftmax" -log_dir = f'{homedir}/logs/{args.expriment_name}/' -im_size = [256, 128] -market_dir = 'Data/Market-1501-v15.09.15' -warmup = args.warmup -m=0.25 -gamma=256 -lamda = args.lamda -num_classes=751 -epoch_size = args.epoch_size -K=args.K -P=args.P -batch_size = K*P -embedding_size=128 -resnet_model_path = '/home/nanshen/xutan/yixin/circlesoftmax/checkpoint/resnet_v2_50.ckpt' # Path to the pretrained model -model_save_path = log_dir + args.modelname # Path to the model.ckpt-(num_steps) will be saved -num_bottleneck = 512 -init_lr = args.init_lr -os.environ["CUDA_VISIBLE_DEVICES"]=args.gpu - - -x = tf.placeholder(tf.float32, [None, im_size[0],im_size[1],3],name='inputs') -y_ = tf.placeholder(tf.int32, [None],name='labels') -lr_backbone = tf.placeholder(tf.float32, name='lr_backbone') -lr_logit = tf.placeholder(tf.float32, name='lr_logit') -is_training = tf.placeholder(tf.bool, name='is_training') -keep_rate = tf.placeholder(tf.float32, name='keep_rate') - - -# 读入数据 -imnum, datasets,le = tf_dataset_PK(P=args.P,K=args.K,erasing_prob=args.erasing_prob,epoch_size=None,path='/home/nanshen/xutan/yixin/market1501/Market-1501-v15.09.15/bounding_box_train') -one_elements=[] -for i in range(751): - iterator = datasets[i].make_one_shot_iterator() - one_element = iterator.get_next() - one_elements.append(one_element) - -def get_PKbatch(sess,one_elements,P=4,K=4): - random.shuffle(one_elements) - this_batch = one_elements[:P] - images,labels = 0,0 - for i in range(P): - if i == 0: - images ,labels = sess.run(this_batch[i]) - else: - image,label = sess.run(this_batch[i]) - images,labels = tf.concat([images,image],0),tf.concat([labels,label],0) - images,labels = sess.run([images,labels]) - return images,labels -# with slim.arg_scope(nets.resnet_v1.resnet_arg_scope()): -# net, endpoints = nets.resnet_v1.resnet_v1_50(x, num_classes=None, -# is_training=is_training,global_pool=True) -# from tf_slim.nets import resnet_v2 -with slim.arg_scope(nets.resnet_v2.resnet_arg_scope()): - net, end_points = nets.resnet_v2.resnet_v2_50(x, num_classes=None, is_training=is_training) - -import math -with tf.variable_scope('Logits'): - # batch_norm_params = {"is_training": is_training, "decay": 0.9, "updates_collections": None} - with slim.arg_scope([slim.conv2d, slim.fully_connected], - weights_regularizer=slim.l2_regularizer(5e-4)): - resnetfeat = tf.squeeze(net, axis=[1, 2]) - fc1 = slim.fully_connected(resnetfeat, num_outputs=num_bottleneck, - activation_fn=None,weights_initializer =tf.random_normal_initializer(mean=0,stddev=math.sqrt(2/num_bottleneck))) - bn1 = slim.batch_norm(fc1, activation_fn=None,is_training=is_training) - drop1 = slim.dropout(bn1, keep_prob=keep_rate) - norm_feat = tf.math.l2_normalize(drop1,axis=1,name='features',epsilon=1e-12) - circle_loss_op,sp,sn,recall,precision,f1_score = PairWiseCircleLoss(norm_feat,y_) - tf.summary.histogram('sn',sn) - tf.summary.histogram('sp',sp) - logits = slim.fully_connected(drop1, num_outputs=num_classes, - activation_fn=None,weights_initializer=tf.random_normal_initializer(mean=0,stddev=0.001)) - -checkpoint_exclude_scopes = 'Logits' -exclusions = None -if checkpoint_exclude_scopes: - exclusions = [ - scope.strip() for scope in checkpoint_exclude_scopes.split(',')] -variables_to_restore = [] -add_var = [] -outof = [] - -for var in slim.get_model_variables(): - excluded = False - for exclusion in exclusions: - if var.op.name.startswith(exclusion): - excluded = True - if not excluded: - variables_to_restore.append(var) - else: - add_var.append(var) -# regularizer = tf.contrib.layers.l2_regularizer(5e-4) -# reg_term = tf.contrib.layers.apply_regularization(regularizer,weights_list=add_var) - -# softmaxloss = tf.nn.sparse_softmax_cross_entropy_with_logits( -# labels=y_, logits=logits) -# softmaxloss = tf.reduce_mean(softmaxloss) -# soft_logit = tf.clip_by_value(soft_logit, 0.000001, 0.999999) -labels = tf.one_hot(y_,num_classes) -soft_logit = tf.nn.softmax(logits) -# softmaxloss = tf.nn.softmax_cross_entropy_with_logits(labels=labels, logits=logits) -soft_logits = tf.clip_by_value(soft_logit, 0.000001, 0.999999) -cross_entropy = -tf.reduce_sum(labels * tf.log(soft_logits), axis=1) -softmaxloss = tf.reduce_mean(cross_entropy) -circle_loss_term = lamda * circle_loss_op/batch_size -loss = softmaxloss + circle_loss_term - - -soft_logit = tf.nn.softmax(logits) -classes = tf.argmax(soft_logit, axis=1, name='classes') -accuracy_softmax = tf.reduce_mean(tf.cast( - tf.equal(tf.cast(classes, dtype=tf.int32), y_), dtype=tf.float32)) - -# optimizer1 = tf.train.AdamOptimizer(learning_rate=lr_backbone) -# optimizer2 = tf.train.AdamOptimizer(learning_rate=lr_logit) -optimizer1 = tf.train.MomentumOptimizer(learning_rate=lr_backbone,momentum=0.9,use_nesterov=True) -optimizer2 = tf.train.MomentumOptimizer(learning_rate=lr_logit,momentum=0.9,use_nesterov=True) - -# bn var -update_ops = tf.get_collection(tf.GraphKeys.UPDATE_OPS) -with tf.control_dependencies(update_ops): - train_step1 = optimizer1.minimize(loss,var_list=variables_to_restore) - train_step2 = optimizer2.minimize(loss,var_list=add_var) - -init = tf.global_variables_initializer() - -saver_restore = tf.train.Saver(var_list=variables_to_restore) -saver = tf.train.Saver(tf.global_variables()) - -config = tf.ConfigProto(allow_soft_placement = True) -config.gpu_options.allow_growth = True #程序按需申请内存 - -# tensorboard记录下来 - -# recall,precision,f1_score -tf.summary.scalar('pairwise_recall',recall) -tf.summary.scalar('pairwise_precision',precision) -tf.summary.scalar('pairwise_f1_score',f1_score) -tf.summary.scalar('softmaxloss',softmaxloss) -# tf.summary.scalar('reg_loss',reg_term) -tf.summary.scalar('circleloss',circle_loss_term) -tf.summary.scalar('totalloss',loss) -tf.summary.scalar('accuracy_softmax',accuracy_softmax) -tf.summary.scalar('lr_backbone',lr_backbone) -tf.summary.scalar('lr_logit',lr_logit) -merged = tf.summary.merge_all() - -def adjust_LR(step,epoch): - lr_l = init_lr * 0.1**(epoch//40) - # if epoch <= 3: - # lr_b = 0 - # else: - # lr_b = 0.1 * lr_l - lr_b = 0.1 * lr_l - if epoch < warmup: - factor = min(1.0, 0.1 + 0.9 / (imnum//batch_size*warmup)) - lr_l,lr_b = factor*lr_l,factor*lr_b - - if step < 2000: # freeze bone trick - lr_b = 0 - - return lr_l,lr_b - -# 开始训练 -with tf.Session(config=config) as sess: - - # writter = tf.summary.FileWriter(log_dir, sess.graph) - train_writer = tf.summary.FileWriter(log_dir + '/train', sess.graph) - # test_writer = tf.summary.FileWriter(log_dir + '/test',sess.graph) - - # 执行初始化操作 - sess.run(init) - - # Load the pretrained checkpoint file xxx.ckpt - saver_restore.restore(sess, resnet_model_path) - - # for i in range(step_size): - # if i % 500 ==0: - # saver.save(sess, log_dir + "/model.ckpt", i) - # if i % 10 == 0: - # # 每10步测试一下 - # summary = sess.run(merged, feed_dict=feed_dict(marketdata,False,x,y_,lr,lr_new)) - # test_writer.add_summary(summary, i) - # else: - # if i % 100 == 99: - # # 每100次记录训练运算时间和内存占用等信息 - # run_options = tf.RunOptions(trace_level=tf.RunOptions.FULL_TRACE) - # run_metadata = tf.RunMetadata() - # summary, _ = sess.run([merged, train_step], - # feed_dict=feed_dict(marketdata,True,x,y_,lr,lr_new), - # options=run_options, - # run_metadata=run_metadata) - # train_writer.add_run_metadata(run_metadata, 'step%03d' % i) - # train_writer.add_summary(summary, i) - # else: - # # 正常情况只进行汇总和训练步伐操作 - # summary, _ = sess.run([merged, train_step], feed_dict=feed_dict(marketdata,True,x,y_,lr,lr_new)) - # train_writer.add_summary(summary, i) - epoch = 0 - sample_counter = 0 - step_counter = 0 - # for i in range(step_size): - import time - t0=time.time() - while epoch < epoch_size: - step_counter+=1 - # epoch = np.mean(traindata._epochs_completed) - import time - ts = time.time() - # images, groundtruth_lists = sess.run(one_element) - images, groundtruth_lists = get_PKbatch(sess,one_elements,P=args.P,K=args.K) - # images_test, groundtruth_lists_test = sess.run(testnext) - - # print(groundtruth_lists) - batch_now = int(images.shape[0]) - sample_counter+=batch_now - epoch = sample_counter // imnum - # if batch_now < batch_size: # skip the last batch - # print('skip last') - # continue - lr_l,lr_b = adjust_LR(step_counter,epoch) - - train_dict = {x: images, - y_: groundtruth_lists, - is_training: True, - lr_logit:lr_l, - lr_backbone:lr_b, - keep_rate:0.5} - # test_dict = {x: images_test, - # y_: groundtruth_lists_test, - # is_training: False, - # lr_logit:lr_l, - # lr_backbone:lr_b, - # keep_rate:1.0} - summary,loss_, acc_,f1,softmaxlossterm,circlelossterm,_,_ = sess.run([merged,loss, accuracy_softmax,f1_score, softmaxloss,circle_loss_term,train_step1,train_step2], feed_dict=train_dict) - # summary_test = sess.run(merged, feed_dict=test_dict) - train_writer.add_summary(summary, step_counter) - # test_writer.add_summary(summary_test,step_counter) - - t1 = time.time() - timingperstep = (t1-ts) - timing = int(t1 - t0) - # train_text = 'Epoch:{}, Step: {}, train_Loss: {:.4f}, train_Accuracy: {:.4f}, Timing:{:.4f}s'.format( - # epoch,step_counter+1, loss_, acc_,t1-t0) - print("Epoch:{} Step:{} TrainLoss:{:.4f} SoftmaxLoss:{:.4f} CircleLoss:{:.4f} SoftmaxAcc:{:.4f} PairwiseF1:{:.4f} TimePerStep:{:.4f}s TotalTiming:{}m{}s".format(epoch,step_counter,loss_,softmaxlossterm,circlelossterm,acc_,f1,timingperstep,(timing)//60,timing%60)) - - if (step_counter+1) % 500 == 0: - saver.save(sess, model_save_path+'.ckpt', global_step=step_counter+1) - tf.io.write_graph(sess.graph, model_save_path, 'graph.pbtxt', as_text=True) - print('save mode to {}'.format(model_save_path)) - - # if (i+1) % 10 == 0: - # t0=time.time() - # images, groundtruth_lists = get_next_batch(marketdata,train=False) - # test_dict = {x: images, - # y_: groundtruth_lists, - # is_training: True, - # lr:lr_new} - # summary,loss_val,acc_val = sess.run([merged,loss, accuracy], feed_dict=test_dict) - # test_writer.add_summary(summary, i) - # t1 = time.time() - # test_text = 'Step: {}, Loss: {:.4f}, Accuracy: {:.4f}, Timing:{:.4f}s'.format( - # i+1, loss_val, acc_val,t1-t0) - # print(test_text) - -train_writer.close() - -os.system('python /home/nanshen/xutan/yixin/circlesoftmax/code/test.py') \ No newline at end of file diff --git a/TensorFlow/contrib/cv/CircleLoss_ID1221_for_TensorFlow/train_gpu.py b/TensorFlow/contrib/cv/CircleLoss_ID1221_for_TensorFlow/train_gpu.py new file mode 100644 index 0000000000000000000000000000000000000000..3135035cfcb1494763af0f2fec4db0b4e9aa13ba --- /dev/null +++ b/TensorFlow/contrib/cv/CircleLoss_ID1221_for_TensorFlow/train_gpu.py @@ -0,0 +1,539 @@ +# Copyright 2017 The TensorFlow Authors. All Rights Reserved. +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. +# ============================================================================ +# Copyright 2021 Huawei Technologies Co., Ltd +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. +from __future__ import absolute_import +from __future__ import division +from __future__ import print_function + +from datetime import datetime +import os.path +import time +import sys +import random +import tensorflow.compat.v1 as tf +tf.disable_eager_execution() +import numpy as np +import importlib +import argparse +import facenet +import lfw +import h5py +import math +from circle_loss import * +from facenet import load_data +from PIL import Image + +def fully_connected(x, units, activation_fn=None,reuse=False): + out=tf.layers.dense(inputs=x,units=units,activation=activation_fn,reuse=reuse) + return out + +def load_and_preprocess_image(img_path): + img_raw = tf.io.read_file(img_path) + img_tensor = tf.cond( + tf.image.is_jpeg(img_raw), + lambda: tf.image.decode_jpeg(img_raw,3), + lambda: tf.image.decode_png(img_raw,3)) + img_tensor = tf.image.resize(img_tensor, [112, 112]) + img_tensor = tf.cast(img_tensor, tf.float32) + img_tensor = (img_tensor - 127.5) / 128.0 + return img_tensor + +def getIteration(args, image_paths_placeholder, labels_placeholder): + image_paths_dataset=tf.data.Dataset.from_tensor_slices(image_paths_placeholder).map(load_and_preprocess_image) + label_dataset=tf.data.Dataset.from_tensor_slices(labels_placeholder) + dataset=tf.data.Dataset.zip((image_paths_dataset,label_dataset)).prefetch( + args.batch_size).batch(args.batch_size) + iteration=dataset.make_initializable_iterator() + return iteration + +def main(args): + network = importlib.import_module(args.model_def) + subdir = datetime.strftime(datetime.now(), '%Y%m%d-%H%M%S') + log_dir = os.path.join(os.path.expanduser(args.logs_base_dir), subdir) + + if not os.path.isdir(log_dir): + os.makedirs(log_dir) + model_dir = os.path.join(os.path.expanduser(args.models_base_dir), subdir) + if not os.path.isdir(model_dir): + os.makedirs(model_dir) + + stat_file_name = os.path.join(log_dir, 'stat.h5') + + facenet.write_arguments_to_file(args, os.path.join(log_dir, 'arguments.txt')) + + src_path, _ = os.path.split(os.path.realpath(__file__)) + facenet.store_revision_info(src_path, log_dir, ' '.join(sys.argv)) + + np.random.seed(seed=args.seed) + random.seed(args.seed) + dataset = facenet.get_dataset(args.data_dir) + if args.filter_filename: + dataset = filter_dataset(dataset, os.path.expanduser(args.filter_filename), + args.filter_percentile, args.filter_min_nrof_images_per_class) + + if args.validation_set_split_ratio > 0.0: + train_set, val_set = facenet.split_dataset(dataset, args.validation_set_split_ratio, + args.min_nrof_val_images_per_class, 'SPLIT_IMAGES') + else: + train_set, val_set = dataset, [] + + nrof_classes = len(train_set) + + print('Model directory: %s' % model_dir) + print('Log directory: %s' % log_dir) + pretrained_model = None + if args.pretrained_model: + pretrained_model = os.path.expanduser(args.pretrained_model) + print('Pre-trained model: %s' % pretrained_model) + + if args.lfw_dir: + print('LFW directory: %s' % args.lfw_dir) + pairs = lfw.read_pairs(os.path.expanduser(args.lfw_pairs)) + lfw_paths, actual_issame = lfw.get_paths(os.path.expanduser(args.lfw_dir), pairs) + + with tf.Graph().as_default(): + tf.set_random_seed(args.seed) + global_step = tf.Variable(0, trainable=False) + + image_list, label_list = facenet.get_image_paths_and_labels(train_set) + assert len(image_list) > 0, 'The training set should not be empty' + + val_image_list, val_label_list = facenet.get_image_paths_and_labels(val_set) + + learning_rate_placeholder = tf.placeholder(tf.float32, name='learning_rate') + phase_train_placeholder = tf.placeholder(tf.bool, name='phase_train') + image_paths_placeholder=tf.placeholder(tf.string,shape=(None,),name="image_paths") + labels_placeholder=tf.placeholder(tf.int32,shape=(None,),name="labels") + + iteration = getIteration(args, image_paths_placeholder, labels_placeholder) + image_batch,label_batch=iteration.get_next() + + image_batch = tf.identity(image_batch, 'image_batch') + image_batch = tf.identity(image_batch, 'input') + label_batch = tf.identity(label_batch, 'label_batch') + + print('Number of classes in training set: %d' % nrof_classes) + print('Number of examples in training set: %d' % len(image_list)) + + print('Number of classes in validation set: %d' % len(val_set)) + print('Number of examples in validation set: %d' % len(val_image_list)) + + print('Building training graph') + + prelogits, _ = network.inference(image_batch, args.keep_probability, + phase_train=phase_train_placeholder, bottleneck_layer_size=args.embedding_size, + weight_decay=args.weight_decay) + logits=fully_connected(prelogits,len(train_set),None,False) + + norm_feat = tf.nn.l2_normalize(logits, 1, 1e-10, name='norm_feat') + + circle_loss_mean = circle_loss(tf.one_hot(label_batch, len(train_set), dtype=tf.float32),norm_feat,batch_size=args.batch_size) + + embeddings = tf.nn.l2_normalize(prelogits, 1, 1e-10, name='embeddings') + + + eps = 1e-4 + prelogits_norm = tf.reduce_mean(tf.norm(tf.abs(prelogits) + eps, ord=args.prelogits_norm_p, axis=1)) + tf.add_to_collection(tf.GraphKeys.REGULARIZATION_LOSSES, prelogits_norm * args.prelogits_norm_loss_factor) + + prelogits_center_loss, _ = facenet.center_loss(prelogits, label_batch, args.center_loss_alfa, nrof_classes) + tf.add_to_collection(tf.GraphKeys.REGULARIZATION_LOSSES, prelogits_center_loss * args.center_loss_factor) + + learning_rate = tf.train.exponential_decay(learning_rate_placeholder, global_step, + args.learning_rate_decay_epochs * args.epoch_size, + args.learning_rate_decay_factor, staircase=True) + tf.summary.scalar('learning_rate', learning_rate) + + tf.add_to_collection('losses', circle_loss_mean) + + correct_prediction = tf.cast(tf.equal(tf.argmax(logits, 1), tf.cast(label_batch, tf.int64)), tf.float32) + accuracy = tf.reduce_mean(correct_prediction) + + regularization_losses = tf.get_collection(tf.GraphKeys.REGULARIZATION_LOSSES) + + total_loss = tf.add_n([circle_loss_mean] + regularization_losses, name='total_loss') + + train_op = facenet.train(total_loss, global_step, args.optimizer, + learning_rate, args.moving_average_decay, tf.global_variables(), args.log_histograms) + + saver = tf.train.Saver(tf.trainable_variables(), max_to_keep=3) + + summary_op = tf.summary.merge_all() + + gpu_options = tf.GPUOptions(per_process_gpu_memory_fraction=args.gpu_memory_fraction) + + sess = tf.Session(config=tf.ConfigProto(gpu_options=gpu_options, log_device_placement=False)) + sess.run(tf.global_variables_initializer()) + sess.run(tf.local_variables_initializer()) + + summary_writer = tf.summary.FileWriter(log_dir, sess.graph) + coord = tf.train.Coordinator() + tf.train.start_queue_runners(coord=coord, sess=sess) + + with sess.as_default(): + + if pretrained_model: + print('Restoring pretrained model: %s' % pretrained_model) + saver.restore(sess, pretrained_model) + + print('Running training') + nrof_steps = args.max_nrof_epochs * args.epoch_size + nrof_val_samples = int(math.ceil( + args.max_nrof_epochs / args.validate_every_n_epochs)) + stat = { + 'loss': np.zeros((nrof_steps,), np.float32), + 'center_loss': np.zeros((nrof_steps,), np.float32), + 'reg_loss': np.zeros((nrof_steps,), np.float32), + 'xent_loss': np.zeros((nrof_steps,), np.float32), + 'prelogits_norm': np.zeros((nrof_steps,), np.float32), + 'accuracy': np.zeros((nrof_steps,), np.float32), + 'val_loss': np.zeros((nrof_val_samples,), np.float32), + 'val_xent_loss': np.zeros((nrof_val_samples,), np.float32), + 'val_accuracy': np.zeros((nrof_val_samples,), np.float32), + 'lfw_accuracy': np.zeros((args.max_nrof_epochs,), np.float32), + 'lfw_valrate': np.zeros((args.max_nrof_epochs,), np.float32), + 'learning_rate': np.zeros((args.max_nrof_epochs,), np.float32), + 'time_train': np.zeros((args.max_nrof_epochs,), np.float32), + 'time_validate': np.zeros((args.max_nrof_epochs,), np.float32), + 'time_evaluate': np.zeros((args.max_nrof_epochs,), np.float32), + 'prelogits_hist': np.zeros((args.max_nrof_epochs, 1000), np.float32), + } + + for epoch in range(1, args.max_nrof_epochs + 1): + step = sess.run(global_step, feed_dict=None) + t = time.time() + + cont = train_epoch(args, sess,epoch, iteration,image_list,label_list,image_paths_placeholder,labels_placeholder, + learning_rate_placeholder, phase_train_placeholder, global_step, + total_loss, train_op, summary_op, summary_writer, regularization_losses, args.learning_rate_schedule_file, + stat, circle_loss_mean, accuracy, + learning_rate, prelogits, prelogits_center_loss, args.random_rotate, args.random_crop, + args.random_flip, prelogits_norm, + args.prelogits_hist_max, args.use_fixed_image_standardization) + + stat['time_train'][epoch - 1] = time.time() - t + if not cont: + break + + save_variables_and_metagraph(sess, saver, summary_writer, model_dir, subdir, epoch) + + t = time.time() + if args.lfw_dir: + acc_mean,acc_var=evaluate(sess,iteration,lfw_paths, image_paths_placeholder, labels_placeholder, phase_train_placeholder, + embeddings,label_batch,actual_issame, args.lfw_batch_size, + args.lfw_nrof_folds, log_dir, step, summary_writer,stat,epoch, + args.lfw_distance_metric, args.lfw_subtract_mean, args.lfw_use_flipped_images, + args.use_fixed_image_standardization) + fout=open("result.txt","a+") + fout.write(str(epoch)+","+str(acc_mean)+"+-"+str(acc_var)+"\n") + fout.close() + stat['time_evaluate'][epoch - 1] = time.time() - t + + print('Saving statistics') + with h5py.File(stat_file_name, 'w') as f: + for key, value in stat.items(): + f.create_dataset(key, data=value) + + return model_dir + +def train_epoch(args,sess,epoch,iteration,image_list,label_list,image_paths_placeholder,labels_placeholder, + learning_rate_placeholder,phase_train_placeholder,step, + loss,train_op,summary_op,summary_writer,reg_losses,learning_rate_schedule_file, + stat,circle_loss_mean,accuracy, + learning_rate,prelogits,prelogits_center_loss,random_rotate,random_crop,random_flip,prelogits_norm, + prelogits_hist_max,use_fixed_image_standardization): + + epoch_idx=np.array(random.sample(range(len(label_list)),args.epoch_size*args.batch_size)) + image_list=np.array(image_list)[epoch_idx] + label_list=np.array(label_list)[epoch_idx] + + batch_number = 0 + if args.learning_rate > 0.0: + lr = args.learning_rate + else: + lr = facenet.get_learning_rate_from_file(learning_rate_schedule_file, epoch) + + if lr <= 0: + return False + feed_dict={image_paths_placeholder:image_list, + labels_placeholder:label_list} + sess.run(iteration.initializer,feed_dict=feed_dict) + + train_time = 0 + while batch_number < args.epoch_size: + start_time = time.time() + + feed_dict = {learning_rate_placeholder: lr, + phase_train_placeholder: True} + + tensor_list = [loss, train_op, step, reg_losses, prelogits, circle_loss_mean, learning_rate, prelogits_norm, + accuracy, prelogits_center_loss] + if batch_number % 100 == 0: + loss_, _, step_, reg_losses_, prelogits_, circle_loss_mean_, lr_, prelogits_norm_, accuracy_, center_loss_, summary_str = sess.run(tensor_list + [summary_op], feed_dict=feed_dict) + summary_writer.add_summary(summary_str, global_step=step_) + else: + loss_, _, step_, reg_losses_, prelogits_, circle_loss_mean_, lr_, prelogits_norm_, accuracy_, center_loss_ = sess.run(tensor_list, feed_dict=feed_dict) + + stat['loss'][step_ - 1] = loss_ + stat['center_loss'][step_ - 1] = center_loss_ + stat['reg_loss'][step_ - 1] = np.sum(reg_losses_) + stat['xent_loss'][step_ - 1] = circle_loss_mean_ + stat['prelogits_norm'][step_ - 1] = prelogits_norm_ + stat['learning_rate'][epoch - 1] = lr_ + stat['accuracy'][step_ - 1] = accuracy_ + stat['prelogits_hist'][epoch - 1, :] += \ + np.histogram(np.minimum(np.abs(prelogits_), prelogits_hist_max), bins=1000, range=(0.0, prelogits_hist_max))[0] + + duration = time.time() - start_time + print( + 'Epoch: [%d][%d/%d]\tTime %.3f\tLoss %2.3f\tXent %2.3f\tRegLoss %2.3f\tAccuracy %2.3f\tLr %2.5f\tCl %2.3f' % + (epoch, batch_number + 1, args.epoch_size, duration, loss_, circle_loss_mean_, np.sum(reg_losses_), + accuracy_, lr_, center_loss_)) + + fout = open("step_time.txt", "a+") + tot_step = (epoch - 1) * 1000 + step_ + fout.write(str(tot_step) + "," + str(duration) + "\n") + fout.close() + + batch_number += 1 + train_time += duration + summary = tf.Summary() + summary.value.add(tag='time/total', simple_value=train_time) + summary_writer.add_summary(summary, global_step=step_) + return True + +def find_threshold(var, percentile): + hist, bin_edges = np.histogram(var, 100) + cdf = np.float32(np.cumsum(hist)) / np.sum(hist) + bin_centers = (bin_edges[:-1] + bin_edges[1:]) / 2 + threshold = np.interp(percentile * 0.01, cdf, bin_centers) + return threshold + + +def filter_dataset(dataset, data_filename, percentile, min_nrof_images_per_class): + with h5py.File(data_filename, 'r') as f: + distance_to_center = np.array(f.get('distance_to_center')) + label_list = np.array(f.get('label_list')) + image_list = np.array(f.get('image_list')) + distance_to_center_threshold = find_threshold(distance_to_center, percentile) + indices = np.where(distance_to_center >= distance_to_center_threshold)[0] + filtered_dataset = dataset + removelist = [] + for i in indices: + label = label_list[i] + image = image_list[i] + if image in filtered_dataset[label].image_paths: + filtered_dataset[label].image_paths.remove(image) + if len(filtered_dataset[label].image_paths) < min_nrof_images_per_class: + removelist.append(label) + + ix = sorted(list(set(removelist)), reverse=True) + for i in ix: + del (filtered_dataset[i]) + + return filtered_dataset + +def evaluate(sess,iteration,lfw_paths,image_paths_placeholder,labels_placeholder, + phase_train_placeholder, + embeddings,label_batch, actual_issame, batch_size, nrof_folds, log_dir, step, summary_writer,stat, epoch, + distance_metric, subtract_mean, use_flipped_images, + use_fixed_image_standardization): + start_time = time.time() + print('Runnning forward pass on LFW images') + + nrof_embeddings = len(actual_issame) * 2 + nrof_images = nrof_embeddings + + embedding_size = int(embeddings.get_shape()[1]) + + assert nrof_images % batch_size == 0, 'The number of LFW images must be an integer multiple of the LFW batch size' + nrof_batches = nrof_images // batch_size + emb_array = np.zeros((nrof_images, embedding_size)) + lab_array = np.zeros((nrof_images,)) + feed_dict={image_paths_placeholder:lfw_paths, + labels_placeholder:np.array(range(len(lfw_paths)))} + sess.run(iteration.initializer,feed_dict=feed_dict) + + for i in range(nrof_batches): + feed_dict = {phase_train_placeholder: False} + emb,lab= sess.run([embeddings,label_batch], feed_dict=feed_dict) + + lab_array[lab] = lab + emb_array[lab, :] = emb + if i % 10 == 9: + print('.', end='') + sys.stdout.flush() + print('') + embeddings = np.zeros((nrof_embeddings, embedding_size)) + if use_flipped_images: + embeddings[:, :embedding_size] = emb_array[0::2, :] + embeddings[:, embedding_size:] = emb_array[1::2, :] + else: + embeddings = emb_array + + assert np.array_equal(lab_array, np.arange( + nrof_images)) == True, 'Wrong labels used for evaluation, possibly caused by training examples left in the input pipeline' + _, _, accuracy, val, val_std, far = lfw.evaluate(embeddings, actual_issame, nrof_folds=nrof_folds, + distance_metric=distance_metric, subtract_mean=subtract_mean) + fout = open("result.txt", "a+") + print('Accuracy: %2.5f+-%2.5f' % (np.mean(accuracy), np.std(accuracy))) + print('Validation rate: %2.5f+-%2.5f @ FAR=%2.5f' % (val, val_std, far)) + lfw_time = time.time() - start_time + summary = tf.Summary() + summary.value.add(tag='lfw/accuracy', simple_value=np.mean(accuracy)) + summary.value.add(tag='lfw/val_rate', simple_value=val) + summary.value.add(tag='time/lfw', simple_value=lfw_time) + summary_writer.add_summary(summary, step) + with open(os.path.join(log_dir, 'lfw_result.txt'), 'at') as f: + f.write('%d\t%.5f\t%.5f\n' % (step, np.mean(accuracy), val)) + stat['lfw_accuracy'][epoch - 1] = np.mean(accuracy) + stat['lfw_valrate'][epoch - 1] = val + return (np.mean(accuracy), np.std(accuracy)) + +def save_variables_and_metagraph(sess, saver, summary_writer, model_dir, model_name, step): + print('Saving variables') + start_time = time.time() + checkpoint_path = os.path.join(model_dir, 'model-%s.ckpt' % model_name) + saver.save(sess, checkpoint_path, global_step=step, write_meta_graph=False) + save_time_variables = time.time() - start_time + print('Variables saved in %.2f seconds' % save_time_variables) + metagraph_filename = os.path.join(model_dir, 'model-%s.meta' % model_name) + save_time_metagraph = 0 + if not os.path.exists(metagraph_filename): + print('Saving metagraph') + start_time = time.time() + saver.export_meta_graph(metagraph_filename) + save_time_metagraph = time.time() - start_time + print('Metagraph saved in %.2f seconds' % save_time_metagraph) + summary = tf.Summary() + summary.value.add(tag='time/save_variables', simple_value=save_time_variables) + summary.value.add(tag='time/save_metagraph', simple_value=save_time_metagraph) + summary_writer.add_summary(summary, step) + + +def parse_arguments(argv): + parser = argparse.ArgumentParser() + parser.add_argument('--logs_base_dir', type=str, + help='Directory where to write event logs.', default='.\\logs') + parser.add_argument('--models_base_dir', type=str, + help='Directory where to write trained models and checkpoints.', default='.\\models') + parser.add_argument('--gpu_memory_fraction', type=float, + help='Upper bound on the amount of GPU memory that will be used by the process.', default=0.9) + parser.add_argument('--pretrained_model', type=str, + help='Load a pretrained model before training starts.',default=".\\20220602-013518\\model-20220602-013518.ckpt-256") + parser.add_argument('--data_dir', type=str, + help='Path to the data directory containing aligned face patches.', + default='D:\\pycode\\dataset\\MS-Celeb-1M_clean_align') + parser.add_argument('--model_def', type=str, + help='Model definition. Points to a module containing the definition of the inference graph.', + default='models.resnet34') + parser.add_argument('--max_nrof_epochs', type=int, + help='Number of epochs to run.', default=1000) + parser.add_argument('--batch_size', type=int, + help='Number of images to process in a batch.', default=30) + parser.add_argument('--image_size', type=int, + help='Image size (height, width) in pixels.', default=112) + parser.add_argument('--epoch_size', type=int, + help='Number of batches per epoch.', default=100) + parser.add_argument('--embedding_size', type=int, + help='Dimensionality of the embedding.', default=512) + parser.add_argument('--random_crop', + help='Performs random cropping of training images. If false, the center image_size pixels from the training images are used. ' + + 'If the size of the images in the data directory is equal to image_size no cropping is performed', + action='store_true') + parser.add_argument('--random_flip', + help='Performs random horizontal flipping of training images.', action='store_true') + parser.add_argument('--random_rotate', + help='Performs random rotations of training images.', action='store_true') + parser.add_argument('--use_fixed_image_standardization', + help='Performs fixed standardization of images.', action='store_true') + parser.add_argument('--keep_probability', type=float, + help='Keep probability of dropout for the fully connected layer(s).', default=1.0) + parser.add_argument('--weight_decay', type=float, + help='L2 weight regularization.', default=0.0) + parser.add_argument('--center_loss_factor', type=float, + help='Center loss factor.', default=0.0) + parser.add_argument('--center_loss_alfa', type=float, + help='Center update rate for center loss.', default=0.95) + parser.add_argument('--prelogits_norm_loss_factor', type=float, + help='Loss based on the norm of the activations in the prelogits layer.', default=0.0) + parser.add_argument('--prelogits_norm_p', type=float, + help='Norm to use for prelogits norm loss.', default=1.0) + parser.add_argument('--prelogits_hist_max', type=float, + help='The max value for the prelogits histogram.', default=10.0) + parser.add_argument('--optimizer', type=str, choices=['ADAGRAD', 'ADADELTA', 'ADAM', 'RMSPROP', 'MOM'], + help='The optimization algorithm to use', default='ADAM') + parser.add_argument('--learning_rate', type=float, + help='Initial learning rate. If set to a negative value a learning rate ' + + 'schedule can be specified in the file "learning_rate_schedule.txt"', default=-1.0) + parser.add_argument('--learning_rate_decay_epochs', type=int, + help='Number of epochs between learning rate decay.', default=300) + parser.add_argument('--learning_rate_decay_factor', type=float, + help='Learning rate decay factor.', default=0.1) + parser.add_argument('--moving_average_decay', type=float, + help='Exponential decay for tracking of training parameters.', default=0.9999) + parser.add_argument('--seed', type=int, + help='Random seed.', default=666) + parser.add_argument('--nrof_preprocess_threads', type=int, + help='Number of preprocessing (data loading and augmentation) threads.', default=4) + parser.add_argument('--log_histograms', + help='Enables logging of weight/bias histograms in tensorboard.', action='store_true') + parser.add_argument('--learning_rate_schedule_file', type=str, + help='File containing the learning rate schedule that is used when learning_rate is set to to -1.', + default='.\\data\\learning_rate_circle_loss.txt') + parser.add_argument('--filter_filename', type=str, + help='File containing image data used for dataset filtering', default='') + parser.add_argument('--filter_percentile', type=float, + help='Keep only the percentile images closed to its class center', default=100.0) + parser.add_argument('--filter_min_nrof_images_per_class', type=int, + help='Keep only the classes with this number of examples or more', default=0) + parser.add_argument('--validate_every_n_epochs', type=int, + help='Number of epoch between validation', default=5) + parser.add_argument('--validation_set_split_ratio', type=float, + help='The ratio of the total dataset to use for validation', default=0.0) + parser.add_argument('--min_nrof_val_images_per_class', type=float, + help='Classes with fewer images will be removed from the validation set', default=0) + + parser.add_argument('--lfw_pairs', type=str, + help='The file containing the pairs to use for validation.', default='.\\data\\pairs.txt') + parser.add_argument('--lfw_dir', type=str, + help='Path to the data directory containing aligned face patches.', default='E:\\dataset\\lfw-deepfunneled_align') + parser.add_argument('--lfw_batch_size', type=int, + help='Number of images to process in a batch in the LFW test set.', default=30) + parser.add_argument('--lfw_nrof_folds', type=int, + help='Number of folds to use for cross validation. Mainly used for testing.', default=10) + parser.add_argument('--lfw_distance_metric', type=int, + help='Type of distance metric to use. 0: Euclidian, 1:Cosine similarity distance.', default=0) + parser.add_argument('--lfw_use_flipped_images', + help='Concatenates embeddings for the image and its horizontally flipped counterpart.', + action='store_true') + parser.add_argument('--lfw_subtract_mean', + help='Subtract feature mean before calculating distance.', action='store_true') + return parser.parse_args(argv) + +if __name__ == '__main__': + main(parse_arguments(sys.argv[1:])) diff --git a/TensorFlow/contrib/cv/CircleLoss_ID1221_for_TensorFlow/train_npu.py b/TensorFlow/contrib/cv/CircleLoss_ID1221_for_TensorFlow/train_npu.py new file mode 100644 index 0000000000000000000000000000000000000000..332ada4e1cf5cba26608eb1e7d657ebc1cb02e45 --- /dev/null +++ b/TensorFlow/contrib/cv/CircleLoss_ID1221_for_TensorFlow/train_npu.py @@ -0,0 +1,542 @@ +# Copyright 2017 The TensorFlow Authors. All Rights Reserved. +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. +# ============================================================================ +# Copyright 2021 Huawei Technologies Co., Ltd +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + +from __future__ import absolute_import +from __future__ import division +from __future__ import print_function +from npu_bridge.npu_init import * + +from datetime import datetime +import os.path +import time +import sys +import random +import tensorflow.compat.v1 as tf +tf.disable_eager_execution() +import numpy as np +import importlib +import argparse +import facenet +import lfw +import h5py +import math +from circle_loss import * +from facenet import load_data +from PIL import Image + +def fully_connected(x, units, activation_fn=None,reuse=False): + out=tf.layers.dense(inputs=x,units=units,activation=activation_fn,reuse=reuse) + return out + +def load_and_preprocess_image(img_path): + img_raw = tf.io.read_file(img_path) + img_tensor = tf.cond( + tf.image.is_jpeg(img_raw), + lambda: tf.image.decode_jpeg(img_raw,3), + lambda: tf.image.decode_png(img_raw,3)) + img_tensor = tf.image.resize(img_tensor, [112, 112]) + img_tensor = tf.cast(img_tensor, tf.float32) + img_tensor = (img_tensor - 127.5) / 128.0 + return img_tensor + +def getIteration(args, image_paths_placeholder, labels_placeholder): + image_paths_dataset=tf.data.Dataset.from_tensor_slices(image_paths_placeholder).map(load_and_preprocess_image) + label_dataset=tf.data.Dataset.from_tensor_slices(labels_placeholder) + dataset=tf.data.Dataset.zip((image_paths_dataset,label_dataset)).prefetch( + args.batch_size).batch(args.batch_size, drop_remainder=True) + iteration=dataset.make_initializable_iterator() + return iteration + +def main(args): + network = importlib.import_module(args.model_def) + subdir = datetime.strftime(datetime.now(), '%Y%m%d-%H%M%S') + log_dir = os.path.join(os.path.expanduser(args.logs_base_dir), subdir) + + if not os.path.isdir(log_dir): + os.makedirs(log_dir) + model_dir = os.path.join(os.path.expanduser(args.models_base_dir), subdir) + if not os.path.isdir(model_dir): + os.makedirs(model_dir) + + stat_file_name = os.path.join(log_dir, 'stat.h5') + + facenet.write_arguments_to_file(args, os.path.join(log_dir, 'arguments.txt')) + + src_path, _ = os.path.split(os.path.realpath(__file__)) + facenet.store_revision_info(src_path, log_dir, ' '.join(sys.argv)) + + np.random.seed(seed=args.seed) + random.seed(args.seed) + dataset = facenet.get_dataset(args.data_dir) + if args.filter_filename: + dataset = filter_dataset(dataset, os.path.expanduser(args.filter_filename), + args.filter_percentile, args.filter_min_nrof_images_per_class) + + if args.validation_set_split_ratio > 0.0: + train_set, val_set = facenet.split_dataset(dataset, args.validation_set_split_ratio, + args.min_nrof_val_images_per_class, 'SPLIT_IMAGES') + else: + train_set, val_set = dataset, [] + + nrof_classes = len(train_set) + + print('Model directory: %s' % model_dir) + print('Log directory: %s' % log_dir) + pretrained_model = None + if args.pretrained_model: + pretrained_model = os.path.expanduser(args.pretrained_model) + print('Pre-trained model: %s' % pretrained_model) + + if args.lfw_dir: + print('LFW directory: %s' % args.lfw_dir) + pairs = lfw.read_pairs(os.path.expanduser(args.lfw_pairs)) + lfw_paths, actual_issame = lfw.get_paths(os.path.expanduser(args.lfw_dir), pairs) + + with tf.Graph().as_default(): + tf.set_random_seed(args.seed) + global_step = tf.Variable(0, trainable=False) + + image_list, label_list = facenet.get_image_paths_and_labels(train_set) + assert len(image_list) > 0, 'The training set should not be empty' + + val_image_list, val_label_list = facenet.get_image_paths_and_labels(val_set) + + learning_rate_placeholder = tf.placeholder(tf.float32, name='learning_rate') + phase_train_placeholder = tf.placeholder(tf.bool, name='phase_train') + image_paths_placeholder=tf.placeholder(tf.string,shape=(None,),name="image_paths") + labels_placeholder=tf.placeholder(tf.int32,shape=(None,),name="labels") + + iteration = getIteration(args, image_paths_placeholder, labels_placeholder) + image_batch,label_batch=iteration.get_next() + + image_batch = tf.identity(image_batch, 'image_batch') + image_batch = tf.identity(image_batch, 'input') + label_batch = tf.identity(label_batch, 'label_batch') + + print('Number of classes in training set: %d' % nrof_classes) + print('Number of examples in training set: %d' % len(image_list)) + + print('Number of classes in validation set: %d' % len(val_set)) + print('Number of examples in validation set: %d' % len(val_image_list)) + + print('Building training graph') + + prelogits, _ = network.inference(image_batch, args.keep_probability, + phase_train=phase_train_placeholder, bottleneck_layer_size=args.embedding_size, + weight_decay=args.weight_decay) + logits=fully_connected(prelogits,len(train_set),None,False) + + norm_feat = tf.nn.l2_normalize(logits, 1, 1e-10, name='norm_feat') + + circle_loss_mean = circle_loss(tf.one_hot(label_batch, len(train_set), dtype=tf.float32),norm_feat,batch_size=args.batch_size) + + embeddings = tf.nn.l2_normalize(prelogits, 1, 1e-10, name='embeddings') + + + eps = 1e-4 + prelogits_norm = tf.reduce_mean(tf.norm(tf.abs(prelogits) + eps, ord=args.prelogits_norm_p, axis=1)) + tf.add_to_collection(tf.GraphKeys.REGULARIZATION_LOSSES, prelogits_norm * args.prelogits_norm_loss_factor) + + prelogits_center_loss, _ = facenet.center_loss(prelogits, label_batch, args.center_loss_alfa, nrof_classes) + tf.add_to_collection(tf.GraphKeys.REGULARIZATION_LOSSES, prelogits_center_loss * args.center_loss_factor) + + learning_rate = tf.train.exponential_decay(learning_rate_placeholder, global_step, + args.learning_rate_decay_epochs * args.epoch_size, + args.learning_rate_decay_factor, staircase=True) + tf.summary.scalar('learning_rate', learning_rate) + + tf.add_to_collection('losses', circle_loss_mean) + + correct_prediction = tf.cast(tf.equal(tf.argmax(logits, 1), tf.cast(label_batch, tf.int64)), tf.float32) + accuracy = tf.reduce_mean(correct_prediction) + + regularization_losses = tf.get_collection(tf.GraphKeys.REGULARIZATION_LOSSES) + + total_loss = tf.add_n([circle_loss_mean] + regularization_losses, name='total_loss') + + train_op = facenet.train(total_loss, global_step, args.optimizer, + learning_rate, args.moving_average_decay, tf.global_variables(), args.log_histograms) + + saver = tf.train.Saver(tf.trainable_variables(), max_to_keep=3) + + summary_op = tf.summary.merge_all() + + gpu_options = tf.GPUOptions(per_process_gpu_memory_fraction=args.gpu_memory_fraction) + + sess = tf.Session(config=npu_config_proto(config_proto=tf.ConfigProto(gpu_options=gpu_options, log_device_placement=False))) + sess.run(tf.global_variables_initializer()) + sess.run(tf.local_variables_initializer()) + + summary_writer = tf.summary.FileWriter(log_dir, sess.graph) + coord = tf.train.Coordinator() + tf.train.start_queue_runners(coord=coord, sess=sess) + + with sess.as_default(): + + if pretrained_model: + print('Restoring pretrained model: %s' % pretrained_model) + saver.restore(sess, pretrained_model) + + print('Running training') + nrof_steps = args.max_nrof_epochs * args.epoch_size + nrof_val_samples = int(math.ceil( + args.max_nrof_epochs / args.validate_every_n_epochs)) + stat = { + 'loss': np.zeros((nrof_steps,), np.float32), + 'center_loss': np.zeros((nrof_steps,), np.float32), + 'reg_loss': np.zeros((nrof_steps,), np.float32), + 'xent_loss': np.zeros((nrof_steps,), np.float32), + 'prelogits_norm': np.zeros((nrof_steps,), np.float32), + 'accuracy': np.zeros((nrof_steps,), np.float32), + 'val_loss': np.zeros((nrof_val_samples,), np.float32), + 'val_xent_loss': np.zeros((nrof_val_samples,), np.float32), + 'val_accuracy': np.zeros((nrof_val_samples,), np.float32), + 'lfw_accuracy': np.zeros((args.max_nrof_epochs,), np.float32), + 'lfw_valrate': np.zeros((args.max_nrof_epochs,), np.float32), + 'learning_rate': np.zeros((args.max_nrof_epochs,), np.float32), + 'time_train': np.zeros((args.max_nrof_epochs,), np.float32), + 'time_validate': np.zeros((args.max_nrof_epochs,), np.float32), + 'time_evaluate': np.zeros((args.max_nrof_epochs,), np.float32), + 'prelogits_hist': np.zeros((args.max_nrof_epochs, 1000), np.float32), + } + + for epoch in range(1, args.max_nrof_epochs + 1): + step = sess.run(global_step, feed_dict=None) + t = time.time() + + cont = train_epoch(args, sess,epoch, iteration,image_list,label_list,image_paths_placeholder,labels_placeholder, + learning_rate_placeholder, phase_train_placeholder, global_step, + total_loss, train_op, summary_op, summary_writer, regularization_losses, args.learning_rate_schedule_file, + stat, circle_loss_mean, accuracy, + learning_rate, prelogits, prelogits_center_loss, args.random_rotate, args.random_crop, + args.random_flip, prelogits_norm, + args.prelogits_hist_max, args.use_fixed_image_standardization) + + stat['time_train'][epoch - 1] = time.time() - t + if not cont: + break + + save_variables_and_metagraph(sess, saver, summary_writer, model_dir, subdir, epoch) + + t = time.time() + if args.lfw_dir: + acc_mean,acc_var=evaluate(sess,iteration,lfw_paths, image_paths_placeholder, labels_placeholder, phase_train_placeholder, + embeddings,label_batch,actual_issame, args.lfw_batch_size, + args.lfw_nrof_folds, log_dir, step, summary_writer,stat,epoch, + args.lfw_distance_metric, args.lfw_subtract_mean, args.lfw_use_flipped_images, + args.use_fixed_image_standardization) + fout=open("result.txt","a+") + fout.write(str(epoch)+","+str(acc_mean)+"+-"+str(acc_var)+"\n") + fout.close() + stat['time_evaluate'][epoch - 1] = time.time() - t + + print('Saving statistics') + with h5py.File(stat_file_name, 'w') as f: + for key, value in stat.items(): + f.create_dataset(key, data=value) + + return model_dir + +def train_epoch(args,sess,epoch,iteration,image_list,label_list,image_paths_placeholder,labels_placeholder, + learning_rate_placeholder,phase_train_placeholder,step, + loss,train_op,summary_op,summary_writer,reg_losses,learning_rate_schedule_file, + stat,circle_loss_mean,accuracy, + learning_rate,prelogits,prelogits_center_loss,random_rotate,random_crop,random_flip,prelogits_norm, + prelogits_hist_max,use_fixed_image_standardization): + + epoch_idx=np.array(random.sample(range(len(label_list)),args.epoch_size*args.batch_size)) + image_list=np.array(image_list)[epoch_idx] + label_list=np.array(label_list)[epoch_idx] + + batch_number = 0 + if args.learning_rate > 0.0: + lr = args.learning_rate + else: + lr = facenet.get_learning_rate_from_file(learning_rate_schedule_file, epoch) + + if lr <= 0: + return False + feed_dict={image_paths_placeholder:image_list, + labels_placeholder:label_list} + sess.run(iteration.initializer,feed_dict=feed_dict) + + train_time = 0 + while batch_number < args.epoch_size: + start_time = time.time() + + feed_dict = {learning_rate_placeholder: lr, + phase_train_placeholder: True} + + tensor_list = [loss, train_op, step, reg_losses, prelogits, circle_loss_mean, learning_rate, prelogits_norm, + accuracy, prelogits_center_loss] + if batch_number % 100 == 0: + loss_, _, step_, reg_losses_, prelogits_, circle_loss_mean_, lr_, prelogits_norm_, accuracy_, center_loss_, summary_str = sess.run(tensor_list + [summary_op], feed_dict=feed_dict) + summary_writer.add_summary(summary_str, global_step=step_) + else: + loss_, _, step_, reg_losses_, prelogits_, circle_loss_mean_, lr_, prelogits_norm_, accuracy_, center_loss_ = sess.run(tensor_list, feed_dict=feed_dict) + + stat['loss'][step_ - 1] = loss_ + stat['center_loss'][step_ - 1] = center_loss_ + stat['reg_loss'][step_ - 1] = np.sum(reg_losses_) + stat['xent_loss'][step_ - 1] = circle_loss_mean_ + stat['prelogits_norm'][step_ - 1] = prelogits_norm_ + stat['learning_rate'][epoch - 1] = lr_ + stat['accuracy'][step_ - 1] = accuracy_ + stat['prelogits_hist'][epoch - 1, :] += \ + np.histogram(np.minimum(np.abs(prelogits_), prelogits_hist_max), bins=1000, range=(0.0, prelogits_hist_max))[0] + + duration = time.time() - start_time + print( + 'Epoch: [%d][%d/%d]\tTime %.3f\tLoss %2.3f\tXent %2.3f\tRegLoss %2.3f\tAccuracy %2.3f\tLr %2.5f\tCl %2.3f' % + (epoch, batch_number + 1, args.epoch_size, duration, loss_, circle_loss_mean_, np.sum(reg_losses_), + accuracy_, lr_, center_loss_)) + + fout = open("step_time.txt", "a+") + tot_step = (epoch - 1) * 1000 + step_ + fout.write(str(tot_step) + "," + str(duration) + "\n") + fout.close() + + batch_number += 1 + train_time += duration + summary = tf.Summary() + summary.value.add(tag='time/total', simple_value=train_time) + summary_writer.add_summary(summary, global_step=step_) + return True + +def find_threshold(var, percentile): + hist, bin_edges = np.histogram(var, 100) + cdf = np.float32(np.cumsum(hist)) / np.sum(hist) + bin_centers = (bin_edges[:-1] + bin_edges[1:]) / 2 + threshold = np.interp(percentile * 0.01, cdf, bin_centers) + return threshold + + +def filter_dataset(dataset, data_filename, percentile, min_nrof_images_per_class): + with h5py.File(data_filename, 'r') as f: + distance_to_center = np.array(f.get('distance_to_center')) + label_list = np.array(f.get('label_list')) + image_list = np.array(f.get('image_list')) + distance_to_center_threshold = find_threshold(distance_to_center, percentile) + indices = np.where(distance_to_center >= distance_to_center_threshold)[0] + filtered_dataset = dataset + removelist = [] + for i in indices: + label = label_list[i] + image = image_list[i] + if image in filtered_dataset[label].image_paths: + filtered_dataset[label].image_paths.remove(image) + if len(filtered_dataset[label].image_paths) < min_nrof_images_per_class: + removelist.append(label) + + ix = sorted(list(set(removelist)), reverse=True) + for i in ix: + del (filtered_dataset[i]) + + return filtered_dataset + +def evaluate(sess,iteration,lfw_paths,image_paths_placeholder,labels_placeholder, + phase_train_placeholder, + embeddings,label_batch, actual_issame, batch_size, nrof_folds, log_dir, step, summary_writer,stat, epoch, + distance_metric, subtract_mean, use_flipped_images, + use_fixed_image_standardization): + start_time = time.time() + print('Runnning forward pass on LFW images') + + nrof_embeddings = len(actual_issame) * 2 + nrof_images = nrof_embeddings + + embedding_size = int(embeddings.get_shape()[1]) + + assert nrof_images % batch_size == 0, 'The number of LFW images must be an integer multiple of the LFW batch size' + nrof_batches = nrof_images // batch_size + emb_array = np.zeros((nrof_images, embedding_size)) + lab_array = np.zeros((nrof_images,)) + feed_dict={image_paths_placeholder:lfw_paths, + labels_placeholder:np.array(range(len(lfw_paths)))} + sess.run(iteration.initializer,feed_dict=feed_dict) + + for i in range(nrof_batches): + feed_dict = {phase_train_placeholder: False} + emb,lab= sess.run([embeddings,label_batch], feed_dict=feed_dict) + + lab_array[lab] = lab + emb_array[lab, :] = emb + if i % 10 == 9: + print('.', end='') + sys.stdout.flush() + print('') + embeddings = np.zeros((nrof_embeddings, embedding_size)) + if use_flipped_images: + embeddings[:, :embedding_size] = emb_array[0::2, :] + embeddings[:, embedding_size:] = emb_array[1::2, :] + else: + embeddings = emb_array + + assert np.array_equal(lab_array, np.arange( + nrof_images)) == True, 'Wrong labels used for evaluation, possibly caused by training examples left in the input pipeline' + _, _, accuracy, val, val_std, far = lfw.evaluate(embeddings, actual_issame, nrof_folds=nrof_folds, + distance_metric=distance_metric, subtract_mean=subtract_mean) + fout = open("result.txt", "a+") + print('Accuracy: %2.5f+-%2.5f' % (np.mean(accuracy), np.std(accuracy))) + print('Validation rate: %2.5f+-%2.5f @ FAR=%2.5f' % (val, val_std, far)) + lfw_time = time.time() - start_time + summary = tf.Summary() + summary.value.add(tag='lfw/accuracy', simple_value=np.mean(accuracy)) + summary.value.add(tag='lfw/val_rate', simple_value=val) + summary.value.add(tag='time/lfw', simple_value=lfw_time) + summary_writer.add_summary(summary, step) + with open(os.path.join(log_dir, 'lfw_result.txt'), 'at') as f: + f.write('%d\t%.5f\t%.5f\n' % (step, np.mean(accuracy), val)) + stat['lfw_accuracy'][epoch - 1] = np.mean(accuracy) + stat['lfw_valrate'][epoch - 1] = val + return (np.mean(accuracy), np.std(accuracy)) + +def save_variables_and_metagraph(sess, saver, summary_writer, model_dir, model_name, step): + print('Saving variables') + start_time = time.time() + checkpoint_path = os.path.join(model_dir, 'model-%s.ckpt' % model_name) + saver.save(sess, checkpoint_path, global_step=step, write_meta_graph=False) + save_time_variables = time.time() - start_time + print('Variables saved in %.2f seconds' % save_time_variables) + metagraph_filename = os.path.join(model_dir, 'model-%s.meta' % model_name) + save_time_metagraph = 0 + if not os.path.exists(metagraph_filename): + print('Saving metagraph') + start_time = time.time() + saver.export_meta_graph(metagraph_filename) + save_time_metagraph = time.time() - start_time + print('Metagraph saved in %.2f seconds' % save_time_metagraph) + summary = tf.Summary() + summary.value.add(tag='time/save_variables', simple_value=save_time_variables) + summary.value.add(tag='time/save_metagraph', simple_value=save_time_metagraph) + summary_writer.add_summary(summary, step) + + +def parse_arguments(argv): + parser = argparse.ArgumentParser() + parser.add_argument('--logs_base_dir', type=str, + help='Directory where to write event logs.', default='./logs') + parser.add_argument('--models_base_dir', type=str, + help='Directory where to write trained models and checkpoints.', default='./models') + parser.add_argument('--gpu_memory_fraction', type=float, + help='Upper bound on the amount of GPU memory that will be used by the process.', default=0.9) + parser.add_argument('--pretrained_model', type=str, + help='Load a pretrained model before training starts.',default="./20220602-013518/model-20220602-013518.ckpt-258") + parser.add_argument('--data_dir', type=str, + help='Path to the data directory containing aligned face patches.', + default='./dataset/MS-Celeb-1M_clean_align') + parser.add_argument('--model_def', type=str, + help='Model definition. Points to a module containing the definition of the inference graph.', + default='models.resnet34') + parser.add_argument('--max_nrof_epochs', type=int, + help='Number of epochs to run.', default=1) + parser.add_argument('--batch_size', type=int, + help='Number of images to process in a batch.', default=30) + parser.add_argument('--image_size', type=int, + help='Image size (height, width) in pixels.', default=112) + parser.add_argument('--epoch_size', type=int, + help='Number of batches per epoch.', default=100) + parser.add_argument('--embedding_size', type=int, + help='Dimensionality of the embedding.', default=512) + parser.add_argument('--random_crop', + help='Performs random cropping of training images. If false, the center image_size pixels from the training images are used. ' + + 'If the size of the images in the data directory is equal to image_size no cropping is performed', + action='store_true') + parser.add_argument('--random_flip', + help='Performs random horizontal flipping of training images.', action='store_true') + parser.add_argument('--random_rotate', + help='Performs random rotations of training images.', action='store_true') + parser.add_argument('--use_fixed_image_standardization', + help='Performs fixed standardization of images.', action='store_true') + parser.add_argument('--keep_probability', type=float, + help='Keep probability of dropout for the fully connected layer(s).', default=1.0) + parser.add_argument('--weight_decay', type=float, + help='L2 weight regularization.', default=0.0) + parser.add_argument('--center_loss_factor', type=float, + help='Center loss factor.', default=0.0) + parser.add_argument('--center_loss_alfa', type=float, + help='Center update rate for center loss.', default=0.95) + parser.add_argument('--prelogits_norm_loss_factor', type=float, + help='Loss based on the norm of the activations in the prelogits layer.', default=0.0) + parser.add_argument('--prelogits_norm_p', type=float, + help='Norm to use for prelogits norm loss.', default=1.0) + parser.add_argument('--prelogits_hist_max', type=float, + help='The max value for the prelogits histogram.', default=10.0) + parser.add_argument('--optimizer', type=str, choices=['ADAGRAD', 'ADADELTA', 'ADAM', 'RMSPROP', 'MOM'], + help='The optimization algorithm to use', default='ADAM') + parser.add_argument('--learning_rate', type=float, + help='Initial learning rate. If set to a negative value a learning rate ' + + 'schedule can be specified in the file "learning_rate_schedule.txt"', default=0.01) + parser.add_argument('--learning_rate_decay_epochs', type=int, + help='Number of epochs between learning rate decay.', default=300) + parser.add_argument('--learning_rate_decay_factor', type=float, + help='Learning rate decay factor.', default=0.1) + parser.add_argument('--moving_average_decay', type=float, + help='Exponential decay for tracking of training parameters.', default=0.9999) + parser.add_argument('--seed', type=int, + help='Random seed.', default=666) + parser.add_argument('--nrof_preprocess_threads', type=int, + help='Number of preprocessing (data loading and augmentation) threads.', default=4) + parser.add_argument('--log_histograms', + help='Enables logging of weight/bias histograms in tensorboard.', action='store_true') + parser.add_argument('--learning_rate_schedule_file', type=str, + help='File containing the learning rate schedule that is used when learning_rate is set to to -1.', + default='./data/learning_rate_circle_loss.txt') + parser.add_argument('--filter_filename', type=str, + help='File containing image data used for dataset filtering', default='') + parser.add_argument('--filter_percentile', type=float, + help='Keep only the percentile images closed to its class center', default=100.0) + parser.add_argument('--filter_min_nrof_images_per_class', type=int, + help='Keep only the classes with this number of examples or more', default=0) + parser.add_argument('--validate_every_n_epochs', type=int, + help='Number of epoch between validation', default=5) + parser.add_argument('--validation_set_split_ratio', type=float, + help='The ratio of the total dataset to use for validation', default=0.0) + parser.add_argument('--min_nrof_val_images_per_class', type=float, + help='Classes with fewer images will be removed from the validation set', default=0) + + parser.add_argument('--lfw_pairs', type=str, + help='The file containing the pairs to use for validation.', default='./data/pairs.txt') + parser.add_argument('--lfw_dir', type=str, + help='Path to the data directory containing aligned face patches.', default='./dataset/lfw-deepfunneled_align') + parser.add_argument('--lfw_batch_size', type=int, + help='Number of images to process in a batch in the LFW test set.', default=30) + parser.add_argument('--lfw_nrof_folds', type=int, + help='Number of folds to use for cross validation. Mainly used for testing.', default=10) + parser.add_argument('--lfw_distance_metric', type=int, + help='Type of distance metric to use. 0: Euclidian, 1:Cosine similarity distance.', default=0) + parser.add_argument('--lfw_use_flipped_images', + help='Concatenates embeddings for the image and its horizontally flipped counterpart.', + action='store_true') + parser.add_argument('--lfw_subtract_mean', + help='Subtract feature mean before calculating distance.', action='store_true') + return parser.parse_args(argv) + +if __name__ == '__main__': + main(parse_arguments(sys.argv[1:])) + diff --git a/TensorFlow/contrib/cv/CircleLoss_ID1221_for_TensorFlow/train_npuv1.py b/TensorFlow/contrib/cv/CircleLoss_ID1221_for_TensorFlow/train_npuv1.py deleted file mode 100644 index ecc73ae406f8f22d00fdfb9f8679496c313c5596..0000000000000000000000000000000000000000 --- a/TensorFlow/contrib/cv/CircleLoss_ID1221_for_TensorFlow/train_npuv1.py +++ /dev/null @@ -1,296 +0,0 @@ -# Copyright 2017 The TensorFlow Authors. All Rights Reserved. -# -# Licensed under the Apache License, Version 2.0 (the "License"); -# you may not use this file except in compliance with the License. -# You may obtain a copy of the License at -# -# http://www.apache.org/licenses/LICENSE-2.0 -# -# Unless required by applicable law or agreed to in writing, software -# distributed under the License is distributed on an "AS IS" BASIS, -# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. -# See the License for the specific language governing permissions and -# limitations under the License. -# ============================================================================ -# Copyright 2021 Huawei Technologies Co., Ltd -# -# Licensed under the Apache License, Version 2.0 (the "License"); -# you may not use this file except in compliance with the License. -# You may obtain a copy of the License at -# -# http://www.apache.org/licenses/LICENSE-2.0 -# -# Unless required by applicable law or agreed to in writing, software -# distributed under the License is distributed on an "AS IS" BASIS, -# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. -# See the License for the specific language governing permissions and -# limitations under the License. - -# from src.dataflow.market import MarketTriplet -# import os -# from npu_bridge.npu_init import * -from npu_bridge.npu_init import * -import numpy as np -import skimage.transform -import tensorflow as tf -# from config import * -# from model import * -slim = tf.contrib.slim -from data import * -from loss import * -from tensorflow.contrib.slim import nets -import moxing as mox -import os - - -import argparse -arg = argparse.ArgumentParser() -# expriment setting -arg.add_argument('-e','--expriment_name',type=str,default='test') -arg.add_argument('-modelname','--modelname',type=str,default='softmaxcircleloss') -arg.add_argument('-epoch','--epoch_size',type=int,default=60) -# arg.add_argument('-g','--gpu',type=str,default='0,1,2,3,4,5,6,7') - -# hyper -arg.add_argument('-w','--warmup',type=int,default=5) -arg.add_argument('-b','--batch_size',type=int,default=32) -arg.add_argument('-lr','--init_lr',type=float,default=0.05) -arg.add_argument('-lamda','--lamda',type=float,default=1) -arg.add_argument('-drop','--droprate',type=float,default=0.5) - -# for huawei npu training -arg.add_argument("--train_url", type=str, default="./output") -arg.add_argument("--data_url", type=str, default="./dataset") -# arg.add_argument("--ckpt_url", type=str, default="./ckpt") -arg.add_argument("--result_dir", type=str, default="/cache/results") -# arg.add_argument("--ckpt_dir", type=str, default="/cache/ckpt") -arg.add_argument("--data_dir", type=str, default="/cache/") -arg.add_argument("--ckptpname", type=str, default="resnet_v1_50.ckpt") - -args = arg.parse_args() - -im_size = [256, 128] -warmup = args.warmup -m=0.25 -gamma=32 -lamda = args.lamda -num_classes=751 -epoch_size = args.epoch_size -batch_size = args.batch_size -init_lr = args.init_lr -num_bottleneck = 512 - -def makedirifnotexist(path): - if not os.path.exists(path): - os.makedirs(path) - - -# backup code -# log_dir -# from tool import copy_code -# copy_code(src="/home/liulizhao/projects/liuyixin/projects",tar=log_dir+"code/") - -# 在ModelArts容器创建数据存放目录 -data_dir = args.data_dir -makedirifnotexist(data_dir) -datasetpath = os.path.join(data_dir,'Market-1501-v15.09.15','bounding_box_train') -# OBS数据拷贝到ModelArts容器内 -mox.file.copy_parallel(args.data_url, data_dir) -# 预训练模型 -# ckpt_dir = args.ckpt_dir -# os.makedirs(ckpt_dir) -# mox.file.copy_parallel(args.ckpt_url, ckpt_dir) -ckptpath = os.path.join(data_dir,'ckpt',args.ckptpname) - -# 在ModelArts容器创建训练输出目录 -result_dir = args.result_dir -makedirifnotexist(result_dir) -# 该实验的目录 -expriment_dir = os.path.join(result_dir,args.expriment_name) -makedirifnotexist(expriment_dir) -# 相关log -log_dir = os.path.join(expriment_dir,'logs') -makedirifnotexist(log_dir) -# 相关ckpt -ckpt_output_dir = os.path.join(expriment_dir,'ckpt') -makedirifnotexist(ckpt_output_dir) - -x = tf.placeholder(tf.float32, [None, im_size[0],im_size[1],3],name='inputs') -y_ = tf.placeholder(tf.int32, [None],name='labels') -lr_backbone = tf.placeholder(tf.float32, name='lr_backbone') -lr_logit = tf.placeholder(tf.float32, name='lr_logit') -is_training = tf.placeholder(tf.bool, name='is_training') -keep_rate = tf.placeholder(tf.float32, name='keep_rate') - - -# 读入数据 -imnum, dataset,le = tf_dataset(batch_size=batch_size,epoch_size=None,path=datasetpath) -print(f'total class num :{len(le.classes_)}') -iterator = dataset.make_one_shot_iterator() -one_element = iterator.get_next() - -with slim.arg_scope(nets.resnet_v1.resnet_arg_scope()): - net, endpoints = nets.resnet_v1.resnet_v1_50(x, num_classes=None, - is_training=is_training,global_pool=True) -import math -with tf.variable_scope('Logits'): - # batch_norm_params = {"is_training": is_training, "decay": 0.9, "updates_collections": None} - with slim.arg_scope([slim.conv2d, slim.fully_connected], - weights_regularizer=slim.l2_regularizer(5e-4)): - resnetfeat = tf.squeeze(net, axis=[1, 2]) - fc1 = slim.fully_connected(resnetfeat, num_outputs=num_bottleneck, - activation_fn=None,weights_initializer =tf.random_normal_initializer(mean=0,stddev=math.sqrt(2/num_bottleneck))) - bn1 = slim.batch_norm(fc1, activation_fn=None,is_training=is_training) - drop1 = slim.dropout(bn1, keep_prob=keep_rate) - norm_feat = tf.math.l2_normalize(drop1,axis=1,name='features',epsilon=1e-12) - logits = slim.fully_connected(drop1, num_outputs=num_classes, - activation_fn=None,weights_initializer=tf.random_normal_initializer(mean=0,stddev=0.001)) - -checkpoint_exclude_scopes = 'Logits' -exclusions = None -if checkpoint_exclude_scopes: - exclusions = [ - scope.strip() for scope in checkpoint_exclude_scopes.split(',')] -variables_to_restore = [] -add_var = [] -outof = [] - -for var in slim.get_model_variables(): - excluded = False - for exclusion in exclusions: - if var.op.name.startswith(exclusion): - excluded = True - if not excluded: - variables_to_restore.append(var) - else: - add_var.append(var) - -softmaxloss = tf.nn.sparse_softmax_cross_entropy_with_logits( - labels=y_, logits=logits) -# softmaxloss = tf.reduce_mean(softmaxloss) -# labels = tf.one_hot(y_,num_classes) -# soft_logit = tf.nn.softmax(logits) -# softmaxloss = tf.nn.softmax_cross_entropy_with_logits(labels=labels, logits=logits) -# soft_logits = tf.clip_by_value(soft_logit, 0.000001, 0.999999) -# cross_entropy = -tf.reduce_sum(labels * tf.log(soft_logits), axis=1) -# softmaxloss = tf.reduce_mean(cross_entropy) -circle_loss_op,sp,sn,recall,precision,f1_score = PairWiseCircleLoss(norm_feat,y_) -circle_loss_term = lamda * circle_loss_op/batch_size -# loss = softmaxloss + circle_loss_term -loss = softmaxloss + circle_loss_term -# loss = softmaxloss - -classes = tf.argmax(logits, axis=1, name='classes') -accuracy_softmax = tf.reduce_mean(tf.cast( - tf.equal(tf.cast(classes, dtype=tf.int32), y_), dtype=tf.float32)) -optimizer1 = tf.train.MomentumOptimizer(learning_rate=lr_backbone,momentum=0.9,use_nesterov=True) -optimizer2 = tf.train.MomentumOptimizer(learning_rate=lr_logit,momentum=0.9,use_nesterov=True) - -# bn var -update_ops = tf.get_collection(tf.GraphKeys.UPDATE_OPS) -with tf.control_dependencies(update_ops): - train_step1 = optimizer1.minimize(loss,var_list=variables_to_restore) - train_step2 = optimizer2.minimize(loss,var_list=add_var) - -init = tf.global_variables_initializer() - -saver_restore = tf.train.Saver(var_list=variables_to_restore) -saver = tf.train.Saver(tf.global_variables()) - -config = tf.ConfigProto() -# config.gpu_options.allow_growth = True #程序按需申请内存 - -# tensorboard记录下来 -# tf.summary.histogram('sn',sn) -# tf.summary.histogram('sp',sp) -# # recall,precision,f1_score -# tf.summary.scalar('pairwise_recall',recall) -# tf.summary.scalar('pairwise_precision',precision) -# tf.summary.scalar('pairwise_f1_score',f1_score) -# tf.summary.scalar('softmaxloss',softmaxloss) -# # tf.summary.scalar('reg_loss',reg_term) -# tf.summary.scalar('circleloss',circle_loss_term) -# tf.summary.scalar('totalloss',loss) -# tf.summary.scalar('accuracy_softmax',accuracy_softmax) -# tf.summary.scalar('lr_backbone',lr_backbone) -# tf.summary.scalar('lr_logit',lr_logit) -# merged = tf.summary.merge_all() - -def adjust_LR(step,epoch): - lr_l = init_lr * 0.1**((epoch)//40) - lr_b = 0.1 * lr_l - if epoch < warmup: - factor = min(1.0, 0.1 + 0.9 / (imnum//batch_size*warmup)) - lr_l,lr_b = factor*lr_l,factor*lr_b - # print(f'学习率: lr_l:{lr_l}, lr_b:{lr_b}!') - return lr_l,lr_b - -# 开始训练 -with tf.Session(config=npu_config_proto(config_proto=config)) as sess: - - # writter = tf.summary.FileWriter(log_dir, sess.graph) - # train_writer = tf.summary.FileWriter(log_dir + '/train', sess.graph) - - # 执行初始化操作 - sess.run(init) - - # Load the pretrained checkpoint file xxx.ckpt - saver_restore.restore(sess, ckptpath) - - epoch = 0 - sample_counter = 0 - step_counter = 0 - # for i in range(step_size): - while epoch < epoch_size: - step_counter+=1 - import time - t0=time.time() - images, groundtruth_lists = sess.run(one_element) - # print(images[0]) - # print(groundtruth_lists) - # print(groundtruth_lists) - batch_now = images.shape[0] - sample_counter+=batch_now - epoch = sample_counter // imnum - if batch_now < batch_size: # skip the last batch - print('skip last') - continue - lr_l,lr_b = adjust_LR(step_counter,epoch) - - train_dict = {x: images, - y_: groundtruth_lists, - is_training: True, - lr_logit:lr_l, - lr_backbone:lr_b, - keep_rate:0.5} - # summary,loss_, acc_,_,_ = sess.run([merged,loss, accuracy_softmax,train_step1,train_step2], feed_dict=train_dict) - sotfmaxloss_,circle_loss_,loss_, acc_, _, _ = sess.run([ softmaxloss,circle_loss_term,loss, accuracy_softmax, train_step1, train_step2], - feed_dict=train_dict) - # sotfmaxloss_,loss_, acc_, _, _ = sess.run([ softmaxloss,loss, accuracy_softmax, train_step1, train_step2], - # feed_dict=train_dict) - # pred,gt = sess.run([classes,y_],feed_dict=train_dict) - # print(f'gt is {gt} and model pred is {pred}!') - # train_writer.add_summary(summary, step_counter) - # train_writer.add_summary(summary, step_counter) - - t1 = time.time() - - train_text = 'Epoch:{}, Step: {}, train_Loss: {:.4f}, train_Accuracy: {:.4f}, Timing:{:.4f}s'.format( - epoch,step_counter+1, loss_, acc_,t1-t0) - print(train_text) - print(f'softmaxloss:{sotfmaxloss_} circleloss:{circle_loss_}') - - if (step_counter+1) % 500 == 0: - saver.save(sess, ckpt_output_dir+'resnet50_circleloss.ckpt', global_step=step_counter+1) - tf.io.write_graph(sess.graph, ckpt_output_dir, 'graph.pbtxt', as_text=True) - print('save mode to {}'.format(ckpt_output_dir)) - - -# 训练结束后，将ModelArts容器内的训练输出拷贝到OBS -mox.file.copy_parallel(result_dir, args.train_url) - -# train_writer.close() - - - - diff --git a/TensorFlow/contrib/cv/CircleLoss_ID1221_for_TensorFlow/train_npuv2.py b/TensorFlow/contrib/cv/CircleLoss_ID1221_for_TensorFlow/train_npuv2.py deleted file mode 100644 index 62936522dc1ec8db10db5b9f3064ec88127a5186..0000000000000000000000000000000000000000 --- a/TensorFlow/contrib/cv/CircleLoss_ID1221_for_TensorFlow/train_npuv2.py +++ /dev/null @@ -1,357 +0,0 @@ -# Copyright 2017 The TensorFlow Authors. All Rights Reserved. -# -# Licensed under the Apache License, Version 2.0 (the "License"); -# you may not use this file except in compliance with the License. -# You may obtain a copy of the License at -# -# http://www.apache.org/licenses/LICENSE-2.0 -# -# Unless required by applicable law or agreed to in writing, software -# distributed under the License is distributed on an "AS IS" BASIS, -# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. -# See the License for the specific language governing permissions and -# limitations under the License. -# ============================================================================ -# Copyright 2021 Huawei Technologies Co., Ltd -# -# Licensed under the Apache License, Version 2.0 (the "License"); -# you may not use this file except in compliance with the License. -# You may obtain a copy of the License at -# -# http://www.apache.org/licenses/LICENSE-2.0 -# -# Unless required by applicable law or agreed to in writing, software -# distributed under the License is distributed on an "AS IS" BASIS, -# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. -# See the License for the specific language governing permissions and -# limitations under the License. - -# from src.dataflow.market import MarketTriplet -import os -# os.system('pip install slim') -from npu_bridge.npu_init import * -import numpy as np -# import skimage.transform -import tensorflow as tf -# from config import * -# from model import * -from data import * -from loss import * -from tensorflow.contrib.slim import nets -import moxing as mox -import os -from tool import copy_code -import argparse -slim = tf.contrib.slim - - - -arg = argparse.ArgumentParser() -# expriment setting -import time -timestamp = int(time.time()) -arg.add_argument('-e','--expriment_name',type=str,default=f'test_{timestamp}') -arg.add_argument('-modelname','--modelname',type=str,default='softmaxcircleloss') -arg.add_argument('-epoch','--epoch_size',type=int,default=60) -# arg.add_argument('-g','--gpu',type=str,default='0,1,2,3,4,5,6,7') - -# hyper -arg.add_argument('-w','--warmup',type=int,default=5) -arg.add_argument('-b','--batch_size',type=int,default=96) -arg.add_argument('-lr','--init_lr',type=float,default=0.05) -arg.add_argument('-lamda','--lamda',type=float,default=1) -arg.add_argument('-drop','--droprate',type=float,default=0.5) - -# for huawei npu training -arg.add_argument("--train_url", type=str, default="./output") -arg.add_argument("--data_url", type=str, default="./dataset") -# arg.add_argument("--ckpt_url", type=str, default="./ckpt") -arg.add_argument("--result_dir", type=str, default="/cache/results/") -# arg.add_argument("--ckpt_dir", type=str, default="/cache/ckpt") -arg.add_argument("--data_dir", type=str, default="/cache/dataset/") -arg.add_argument("--ckptpname", type=str, default="resnet_v1_50.ckpt") - -args = arg.parse_args() - - -im_size = [256, 128] -warmup = args.warmup -m=0.25 -gamma=32 -lamda = args.lamda -num_classes=751 -epoch_size = args.epoch_size -batch_size = args.batch_size -init_lr = args.init_lr -num_bottleneck = 512 - -def makedirifnotexist(path): - if not os.path.exists(path): - os.makedirs(path) - -x = tf.placeholder(tf.float32, [None, im_size[0],im_size[1],3],name='inputs') -y_ = tf.placeholder(tf.int32, [None],name='labels') -lr_backbone = tf.placeholder(tf.float32, name='lr_backbone') -lr_logit = tf.placeholder(tf.float32, name='lr_logit') -is_training = tf.placeholder(tf.bool, name='is_training') -keep_rate = tf.placeholder(tf.float32, name='keep_rate') - -# OBS数据拷贝到ModelArts容器内 -data_dir = args.data_dir -mox.file.copy_parallel(args.data_url, data_dir) - - -# 在ModelArts容器创建数据存放目录 -makedirifnotexist(data_dir) -datasetpath = os.path.join(data_dir,'bounding_box_train') - -# 预训练模型 -ckptpath = os.path.join(data_dir,args.ckptpname) - -# 在ModelArts容器创建训练输出目录 -result_dir = args.result_dir -makedirifnotexist(result_dir) -# 该实验的目录 -expriment_dir = os.path.join(result_dir,args.expriment_name) -makedirifnotexist(expriment_dir) -# 相关log -log_dir = os.path.join(expriment_dir,'logs') -makedirifnotexist(log_dir) -# 相关ckpt -ckpt_output_dir = os.path.join(expriment_dir,'ckpt') -makedirifnotexist(ckpt_output_dir) -model_save_path = ckpt_output_dir + args.expriment_name - -x = tf.placeholder(tf.float32, [None, im_size[0],im_size[1],3],name='inputs') -y_ = tf.placeholder(tf.int32, [None],name='labels') -lr_backbone = tf.placeholder(tf.float32, name='lr_backbone') -lr_logit = tf.placeholder(tf.float32, name='lr_logit') -is_training = tf.placeholder(tf.bool, name='is_training') -keep_rate = tf.placeholder(tf.float32, name='keep_rate') - - -# 读入数据 -imnum, dataset,le = tf_dataset(batch_size=batch_size,epoch_size=None,path=datasetpath) -print(f'total class num :{len(le.classes_)}') -iterator = dataset.make_one_shot_iterator() -one_element = iterator.get_next() - -# valdataset = tf_val_dataset(batch_size=batch_size,epoch_size=None) -# testiterator = valdataset.make_one_shot_iterator() -# testnext = testiterator.get_next() - -# 定义计算图 -# y_logit = None -# with slim.arg_scope(resnet_arg_scope(is_training=False)): -# y_logit = resnet_v2_50(x, num_classes=embedding_size,global_pool=True) -# y_norm = tf.math.l2_normalize(y_logit,axis=1,name='features',epsilon=1e-12) -# loss_op,sn,sp = get_loss_and_metric(y_norm,y_) -# train_step = tf.train.AdamOptimizer(lr).minimize(loss_op) - - -with slim.arg_scope(nets.resnet_v1.resnet_arg_scope()): - net, endpoints = nets.resnet_v1.resnet_v1_50(x, num_classes=None, - is_training=is_training,global_pool=True) -import math -with tf.variable_scope('Logits'): - # batch_norm_params = {"is_training": is_training, "decay": 0.9, "updates_collections": None} - with slim.arg_scope([slim.conv2d, slim.fully_connected], - weights_regularizer=slim.l2_regularizer(5e-4)): - resnetfeat = tf.squeeze(net, axis=[1, 2]) - fc1 = slim.fully_connected(resnetfeat, num_outputs=num_bottleneck, - activation_fn=None,weights_initializer =tf.random_normal_initializer(mean=0,stddev=math.sqrt(2/num_bottleneck))) - bn1 = slim.batch_norm(fc1, activation_fn=None,is_training=is_training) - drop1 = slim.dropout(bn1, keep_prob=keep_rate) - norm_feat = tf.math.l2_normalize(drop1,axis=1,name='features',epsilon=1e-12) - circle_loss_op,sp,sn,recall,precision,f1_score = PairWiseCircleLoss(norm_feat,y_) - tf.summary.histogram('sn',sn) - tf.summary.histogram('sp',sp) - logits = slim.fully_connected(drop1, num_outputs=num_classes, - activation_fn=None,weights_initializer=tf.random_normal_initializer(mean=0,stddev=0.001)) - -checkpoint_exclude_scopes = 'Logits' -exclusions = None -if checkpoint_exclude_scopes: - exclusions = [ - scope.strip() for scope in checkpoint_exclude_scopes.split(',')] -variables_to_restore = [] -add_var = [] -outof = [] - -for var in slim.get_model_variables(): - excluded = False - for exclusion in exclusions: - if var.op.name.startswith(exclusion): - excluded = True - if not excluded: - variables_to_restore.append(var) - else: - add_var.append(var) -# regularizer = tf.contrib.layers.l2_regularizer(5e-4) -# reg_term = tf.contrib.layers.apply_regularization(regularizer,weights_list=add_var) - -# softmaxloss = tf.nn.sparse_softmax_cross_entropy_with_logits( -# labels=y_, logits=logits) -# softmaxloss = tf.reduce_mean(softmaxloss) -# soft_logit = tf.clip_by_value(soft_logit, 0.000001, 0.999999) -labels = tf.one_hot(y_,num_classes) -soft_logit = tf.nn.softmax(logits) -# softmaxloss = tf.nn.softmax_cross_entropy_with_logits(labels=labels, logits=logits) -soft_logits = tf.clip_by_value(soft_logit, 0.000001, 0.999999) -cross_entropy = -tf.reduce_sum(labels * tf.log(soft_logits), axis=1) -softmaxloss = tf.reduce_mean(cross_entropy) -circle_loss_term = lamda * circle_loss_op/batch_size -loss = softmaxloss + circle_loss_term - - -soft_logit = tf.nn.softmax(logits) -classes = tf.argmax(soft_logit, axis=1, name='classes') -accuracy_softmax = tf.reduce_mean(tf.cast( - tf.equal(tf.cast(classes, dtype=tf.int32), y_), dtype=tf.float32)) - -# optimizer1 = tf.train.AdamOptimizer(learning_rate=lr_backbone) -# optimizer2 = tf.train.AdamOptimizer(learning_rate=lr_logit) -optimizer1 = tf.train.MomentumOptimizer(learning_rate=lr_backbone,momentum=0.9,use_nesterov=True) -optimizer2 = tf.train.MomentumOptimizer(learning_rate=lr_logit,momentum=0.9,use_nesterov=True) - -# bn var -update_ops = tf.get_collection(tf.GraphKeys.UPDATE_OPS) -with tf.control_dependencies(update_ops): - train_step1 = optimizer1.minimize(loss,var_list=variables_to_restore) - train_step2 = optimizer2.minimize(loss,var_list=add_var) - -init = tf.global_variables_initializer() - -saver_restore = tf.train.Saver(var_list=variables_to_restore) -saver = tf.train.Saver(tf.global_variables()) - -config = tf.ConfigProto(allow_soft_placement = True) -config.gpu_options.allow_growth = True #程序按需申请内存 - -# tensorboard记录下来 - -# recall,precision,f1_score -tf.summary.scalar('pairwise_recall',recall) -tf.summary.scalar('pairwise_precision',precision) -tf.summary.scalar('pairwise_f1_score',f1_score) -tf.summary.scalar('softmaxloss',softmaxloss) -# tf.summary.scalar('reg_loss',reg_term) -tf.summary.scalar('circleloss',circle_loss_term) -tf.summary.scalar('totalloss',loss) -tf.summary.scalar('accuracy_softmax',accuracy_softmax) -tf.summary.scalar('lr_backbone',lr_backbone) -tf.summary.scalar('lr_logit',lr_logit) -merged = tf.summary.merge_all() - -def adjust_LR(step,epoch): - lr_l = init_lr * 0.1**((epoch)//40) - lr_b = 0.1 * lr_l - if epoch < warmup: - factor = min(1.0, 0.1 + 0.9 / (imnum//batch_size*warmup)) - lr_l,lr_b = factor*lr_l,factor*lr_b - return lr_l,lr_b - -# 开始训练 -with tf.Session(config=npu_config_proto(config_proto=config)) as sess: - - # writter = tf.summary.FileWriter(log_dir, sess.graph) - train_writer = tf.summary.FileWriter(log_dir + '/train', sess.graph) - - # 执行初始化操作 - sess.run(init) - - # Load the pretrained checkpoint file xxx.ckpt - saver_restore.restore(sess, ckptpath) - - # for i in range(step_size): - # if i % 500 ==0: - # saver.save(sess, log_dir + "/model.ckpt", i) - # if i % 10 == 0: - # # 每10步测试一下 - # summary = sess.run(merged, feed_dict=feed_dict(marketdata,False,x,y_,lr,lr_new)) - # test_writer.add_summary(summary, i) - # else: - # if i % 100 == 99: - # # 每100次记录训练运算时间和内存占用等信息 - # run_options = tf.RunOptions(trace_level=tf.RunOptions.FULL_TRACE) - # run_metadata = tf.RunMetadata() - # summary, _ = sess.run([merged, train_step], - # feed_dict=feed_dict(marketdata,True,x,y_,lr,lr_new), - # options=run_options, - # run_metadata=run_metadata) - # train_writer.add_run_metadata(run_metadata, 'step%03d' % i) - # train_writer.add_summary(summary, i) - # else: - # # 正常情况只进行汇总和训练步伐操作 - # summary, _ = sess.run([merged, train_step], feed_dict=feed_dict(marketdata,True,x,y_,lr,lr_new)) - # train_writer.add_summary(summary, i) - epoch = 0 - sample_counter = 0 - step_counter = 0 - # for i in range(step_size): - import time - t0=time.time() - while epoch < epoch_size: - step_counter+=1 - # epoch = np.mean(traindata._epochs_completed) - import time - ts = time.time() - - images, groundtruth_lists = sess.run(one_element) - # images_test, groundtruth_lists_test = sess.run(testnext) - - # print(groundtruth_lists) - batch_now = images.shape[0] - sample_counter+=batch_now - epoch = sample_counter // imnum - if batch_now < batch_size: # skip the last batch - print('skip last') - continue - lr_l,lr_b = adjust_LR(step_counter,epoch) - - train_dict = {x: images, - y_: groundtruth_lists, - is_training: True, - lr_logit:lr_l, - lr_backbone:lr_b, - keep_rate:0.5} - # test_dict = {x: images_test, - # y_: groundtruth_lists_test, - # is_training: False, - # lr_logit:lr_l, - # lr_backbone:lr_b, - # keep_rate:1.0} - summary,loss_, acc_,f1,softmaxlossterm,circlelossterm,_,_ = sess.run([merged,loss, accuracy_softmax,f1_score, softmaxloss,circle_loss_term,train_step1,train_step2], feed_dict=train_dict) - # summary_test = sess.run(merged, feed_dict=test_dict) - train_writer.add_summary(summary, step_counter) - # test_writer.add_summary(summary_test,step_counter) - - t1 = time.time() - timing = int(t1 - t0) - timingperstep = (t1-ts) - # train_text = 'Epoch:{}, Step: {}, train_Loss: {:.4f}, train_Accuracy: {:.4f}, Timing:{:.4f}s'.format( - # epoch,step_counter+1, loss_, acc_,t1-t0) - print("Epoch:{} Step:{} TrainLoss:{:.4f} SoftmaxLoss:{:.4f} CircleLoss:{:.4f} SoftmaxAcc:{:.4f} PairwiseF1:{:.4f} TimePerStep:{:.4f}s TotalTiming:{}m{}s".format(epoch,step_counter,loss_,softmaxlossterm,circlelossterm,acc_,f1,timingperstep,(timing)//60,timing%60)) - - if (step_counter+1) % 500 == 0: - saver.save(sess, model_save_path+'.ckpt', global_step=step_counter+1) - tf.io.write_graph(sess.graph, model_save_path, 'graph.pbtxt', as_text=True) - print('save mode to {}'.format(model_save_path)) - - # if (i+1) % 10 == 0: - # t0=time.time() - # images, groundtruth_lists = get_next_batch(marketdata,train=False) - # test_dict = {x: images, - # y_: groundtruth_lists, - # is_training: True, - # lr:lr_new} - # summary,loss_val,acc_val = sess.run([merged,loss, accuracy], feed_dict=test_dict) - # test_writer.add_summary(summary, i) - # t1 = time.time() - # test_text = 'Step: {}, Loss: {:.4f}, Accuracy: {:.4f}, Timing:{:.4f}s'.format( - # i+1, loss_val, acc_val,t1-t0) - # print(test_text) - -train_writer.close() -# 训练结束后，将ModelArts容器内的训练输出拷贝到OBS -mox.file.copy_parallel(result_dir, args.train_url) \ No newline at end of file diff --git a/TensorFlow/contrib/cv/CircleLoss_ID1221_for_TensorFlow/train_npuv3.py b/TensorFlow/contrib/cv/CircleLoss_ID1221_for_TensorFlow/train_npuv3.py deleted file mode 100644 index 07dc001d932db9aedd41955937c2c7ec0f4f256d..0000000000000000000000000000000000000000 --- a/TensorFlow/contrib/cv/CircleLoss_ID1221_for_TensorFlow/train_npuv3.py +++ /dev/null @@ -1,336 +0,0 @@ -# Copyright 2017 The TensorFlow Authors. All Rights Reserved. -# -# Licensed under the Apache License, Version 2.0 (the "License"); -# you may not use this file except in compliance with the License. -# You may obtain a copy of the License at -# -# http://www.apache.org/licenses/LICENSE-2.0 -# -# Unless required by applicable law or agreed to in writing, software -# distributed under the License is distributed on an "AS IS" BASIS, -# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. -# See the License for the specific language governing permissions and -# limitations under the License. -# ============================================================================ -# Copyright 2021 Huawei Technologies Co., Ltd -# -# Licensed under the Apache License, Version 2.0 (the "License"); -# you may not use this file except in compliance with the License. -# You may obtain a copy of the License at -# -# http://www.apache.org/licenses/LICENSE-2.0 -# -# Unless required by applicable law or agreed to in writing, software -# distributed under the License is distributed on an "AS IS" BASIS, -# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. -# See the License for the specific language governing permissions and -# limitations under the License. - - -import argparse -import time -import os - - -def parse_args(): - timestamp = int(time.time()) - parser = argparse.ArgumentParser(description='MAIN') - parser.add_argument('-b','--batch_size',type=int,default=64) - parser.add_argument('-epoch','--epoch_size',type=int,default=120) - parser.add_argument('-w','--warmup',type=int,default=5) - parser.add_argument('--init_lr',type=float,default=0.08) - parser.add_argument('--lamda',type=float,default=1) - parser.add_argument('--data_path', type=str, default="./dataset/") - parser.add_argument('--result_dir', type=str, default="./result/") - parser.add_argument('--ckptpname', type=str, default="resnet_v1_50.ckpt") - parser.add_argument('-e','--expriment_name',type=str,default=f'test_{timestamp}') - args = parser.parse_args() - return args - -args = parse_args() - - -m=0.25 -gamma=32 -num_classes=751 -num_bottleneck = 512 -im_size = [256, 128] -lamda = args.lamda -warmup = args.warmup -epoch_size = args.epoch_size -batch_size = args.batch_size -init_lr = args.init_lr -data_dir = args.data_path -ckptpname = args.ckptpname -result_dir = args.result_dir -expriment_name = args.expriment_name - - - -if __name__ == '__main__': - from npu_bridge.npu_init import * - import numpy as np - import tensorflow as tf - from data import * - from loss import * - from tensorflow.contrib.slim import nets - from tool import copy_code - - slim = tf.contrib.slim - - - def makedirifnotexist(path): - if not os.path.exists(path): - os.makedirs(path) - - datasetpath = os.path.join(data_dir, 'bounding_box_train') - print(datasetpath) - ckptpath = os.path.join(data_dir,ckptpname) - expriment_dir = os.path.join(result_dir,expriment_name) - log_dir = os.path.join(expriment_dir, 'logs') - ckpt_output_dir = os.path.join(expriment_dir, 'ckpt') - model_save_path = ckpt_output_dir + expriment_name - - makedirifnotexist(result_dir) - makedirifnotexist(expriment_dir) - makedirifnotexist(log_dir) - makedirifnotexist(ckpt_output_dir) - - x = tf.placeholder(tf.float32, [None, im_size[0],im_size[1],3],name='inputs') - y_ = tf.placeholder(tf.int32, [None],name='labels') - lr_backbone = tf.placeholder(tf.float32, name='lr_backbone') - lr_logit = tf.placeholder(tf.float32, name='lr_logit') - is_training = tf.placeholder(tf.bool, name='is_training') - keep_rate = tf.placeholder(tf.float32, name='keep_rate') - - # 读入数据 - imnum, dataset,le = tf_dataset(batch_size=batch_size,epoch_size=None,path=datasetpath) - print(f'total class num :{len(le.classes_)}') - iterator = dataset.make_one_shot_iterator() - one_element = iterator.get_next() - - # valdataset = tf_val_dataset(batch_size=batch_size,epoch_size=None) - # testiterator = valdataset.make_one_shot_iterator() - # testnext = testiterator.get_next() - - # 定义计算图 - # y_logit = None - # with slim.arg_scope(resnet_arg_scope(is_training=False)): - # y_logit = resnet_v2_50(x, num_classes=embedding_size,global_pool=True) - # y_norm = tf.math.l2_normalize(y_logit,axis=1,name='features',epsilon=1e-12) - # loss_op,sn,sp = get_loss_and_metric(y_norm,y_) - # train_step = tf.train.AdamOptimizer(lr).minimize(loss_op) - - - with slim.arg_scope(nets.resnet_v1.resnet_arg_scope()): - net, endpoints = nets.resnet_v1.resnet_v1_50(x, num_classes=None, - is_training=is_training,global_pool=True) - import math - # with slim.arg_scope(nets.resnet_v2.resnet_arg_scope()): - # net, end_points = nets.resnet_v2.resnet_v2_50(x, num_classes=None, is_training=is_training) - - with tf.variable_scope('Logits'): - # batch_norm_params = {"is_training": is_training, "decay": 0.9, "updates_collections": None} - with slim.arg_scope([slim.conv2d, slim.fully_connected], - weights_regularizer=slim.l2_regularizer(5e-4)): - resnetfeat = tf.squeeze(net, axis=[1, 2]) - fc1 = slim.fully_connected(resnetfeat, num_outputs=num_bottleneck, - activation_fn=None,weights_initializer =tf.random_normal_initializer(mean=0,stddev=math.sqrt(2/num_bottleneck))) - bn1 = slim.batch_norm(fc1, activation_fn=None,is_training=is_training) - drop1 = slim.dropout(bn1, keep_prob=keep_rate) - norm_feat = tf.math.l2_normalize(drop1,axis=1,name='features',epsilon=1e-12) - circle_loss_op,sp,sn,recall,precision,f1_score = PairWiseCircleLoss(norm_feat,y_) - tf.summary.histogram('sn',sn) - tf.summary.histogram('sp',sp) - logits = slim.fully_connected(drop1, num_outputs=num_classes,activation_fn=None,weights_initializer=tf.random_normal_initializer(mean=0,stddev=0.001)) - - checkpoint_exclude_scopes = 'Logits' - exclusions = None - if checkpoint_exclude_scopes: - exclusions = [ - scope.strip() for scope in checkpoint_exclude_scopes.split(',')] - variables_to_restore = [] - add_var = [] - outof = [] - - for var in slim.get_model_variables(): - excluded = False - for exclusion in exclusions: - if var.op.name.startswith(exclusion): - excluded = True - if not excluded: - variables_to_restore.append(var) - else: - add_var.append(var) - # regularizer = tf.contrib.layers.l2_regularizer(5e-4) - # reg_term = tf.contrib.layers.apply_regularization(regularizer,weights_list=add_var) - - # softmaxloss = tf.nn.sparse_softmax_cross_entropy_with_logits( - # labels=y_, logits=logits) - # softmaxloss = tf.reduce_mean(softmaxloss) - # soft_logit = tf.clip_by_value(soft_logit, 0.000001, 0.999999) - labels = tf.one_hot(y_,num_classes) - soft_logit = tf.nn.softmax(logits) - # softmaxloss = tf.nn.softmax_cross_entropy_with_logits(labels=labels, logits=logits) - soft_logits = tf.clip_by_value(soft_logit, 0.000001, 0.999999) - cross_entropy = -tf.reduce_sum(labels * tf.log(soft_logits), axis=1) - softmaxloss = tf.reduce_mean(cross_entropy) - circle_loss_term = lamda * circle_loss_op/batch_size - loss = softmaxloss + circle_loss_term - - - soft_logit = tf.nn.softmax(logits) - classes = tf.argmax(soft_logit, axis=1, name='classes') - accuracy_softmax = tf.reduce_mean(tf.cast( - tf.equal(tf.cast(classes, dtype=tf.int32), y_), dtype=tf.float32)) - - # optimizer1 = tf.train.AdamOptimizer(learning_rate=lr_backbone) - # optimizer2 = tf.train.AdamOptimizer(learning_rate=lr_logit) - optimizer1 = tf.train.MomentumOptimizer(learning_rate=lr_backbone,momentum=0.9,use_nesterov=True) - optimizer2 = tf.train.MomentumOptimizer(learning_rate=lr_logit,momentum=0.9,use_nesterov=True) - - # bn var - update_ops = tf.get_collection(tf.GraphKeys.UPDATE_OPS) - with tf.control_dependencies(update_ops): - train_step1 = optimizer1.minimize(loss,var_list=variables_to_restore) - train_step2 = optimizer2.minimize(loss,var_list=add_var) - - init = tf.global_variables_initializer() - - saver_restore = tf.train.Saver(var_list=variables_to_restore) - saver = tf.train.Saver(tf.global_variables()) - - config = tf.ConfigProto(allow_soft_placement = True) - config.gpu_options.allow_growth = True #程序按需申请内存 - - # tensorboard记录下来 - - # recall,precision,f1_score - tf.summary.scalar('pairwise_recall',recall) - tf.summary.scalar('pairwise_precision',precision) - tf.summary.scalar('pairwise_f1_score',f1_score) - tf.summary.scalar('softmaxloss',softmaxloss) - # tf.summary.scalar('reg_loss',reg_term) - tf.summary.scalar('circleloss',circle_loss_term) - tf.summary.scalar('totalloss',loss) - tf.summary.scalar('accuracy_softmax',accuracy_softmax) - tf.summary.scalar('lr_backbone',lr_backbone) - tf.summary.scalar('lr_logit',lr_logit) - merged = tf.summary.merge_all() - - - def adjust_LR(step,epoch): - lr_l = init_lr * 0.1**((epoch)//40) - lr_b = 0.1 * lr_l - if epoch < warmup: - factor = min(1.0, 0.1 + 0.9 / (imnum//batch_size*warmup)) - lr_l,lr_b = factor*lr_l,factor*lr_b - if step < 32000//args.batch_size: # freeze bonee - lr_b = 0 - return lr_l,lr_b - - # 开始训练 - with tf.Session(config=npu_config_proto(config_proto=config)) as sess: - - # writter = tf.summary.FileWriter(log_dir, sess.graph) - train_writer = tf.summary.FileWriter(log_dir + '/train', sess.graph) - - # 执行初始化操作 - sess.run(init) - - # Load the pretrained checkpoint file xxx.ckpt - saver_restore.restore(sess, ckptpath) - - # for i in range(step_size): - # if i % 500 ==0: - # saver.save(sess, log_dir + "/model.ckpt", i) - # if i % 10 == 0: - # # 每10步测试一下 - # summary = sess.run(merged, feed_dict=feed_dict(marketdata,False,x,y_,lr,lr_new)) - # test_writer.add_summary(summary, i) - # else: - # if i % 100 == 99: - # # 每100次记录训练运算时间和内存占用等信息 - # run_options = tf.RunOptions(trace_level=tf.RunOptions.FULL_TRACE) - # run_metadata = tf.RunMetadata() - # summary, _ = sess.run([merged, train_step], - # feed_dict=feed_dict(marketdata,True,x,y_,lr,lr_new), - # options=run_options, - # run_metadata=run_metadata) - # train_writer.add_run_metadata(run_metadata, 'step%03d' % i) - # train_writer.add_summary(summary, i) - # else: - # # 正常情况只进行汇总和训练步伐操作 - # summary, _ = sess.run([merged, train_step], feed_dict=feed_dict(marketdata,True,x,y_,lr,lr_new)) - # train_writer.add_summary(summary, i) - epoch = 0 - sample_counter = 0 - step_counter = 0 - # for i in range(step_size): - import time - t0=time.time() - while epoch < epoch_size: - step_counter+=1 - # epoch = np.mean(traindata._epochs_completed) - import time - ts = time.time() - images, groundtruth_lists = sess.run(one_element) - # images_test, groundtruth_lists_test = sess.run(testnext) - - # print(groundtruth_lists) - batch_now = images.shape[0] - sample_counter+=batch_now - epoch = sample_counter // imnum - if batch_now < batch_size: # skip the last batch - print('skip last') - continue - lr_l,lr_b = adjust_LR(step_counter,epoch) - - train_dict = {x: images, - y_: groundtruth_lists, - is_training: True, - lr_logit:lr_l, - lr_backbone:lr_b, - keep_rate:0.5} - # test_dict = {x: images_test, - # y_: groundtruth_lists_test, - # is_training: False, - # lr_logit:lr_l, - # lr_backbone:lr_b, - # keep_rate:1.0} - summary,loss_, acc_,f1,softmaxlossterm,circlelossterm,_,_ = sess.run([merged,loss, accuracy_softmax,f1_score, softmaxloss,circle_loss_term,train_step1,train_step2], feed_dict=train_dict) - # summary_test = sess.run(merged, feed_dict=test_dict) - train_writer.add_summary(summary, step_counter) - # test_writer.add_summary(summary_test,step_counter) - - t1 = time.time() - timingperstep = (t1 - ts) - timing = int(t1 - t0) - # train_text = 'Epoch:{}, Step: {}, train_Loss: {:.4f}, train_Accuracy: {:.4f}, Timing:{:.4f}s'.format( - # epoch,step_counter+1, loss_, acc_,t1-t0) - print( - "Epoch:{} Step:{} TrainLoss:{:.4f} SoftmaxLoss:{:.4f} CircleLoss:{:.4f} SoftmaxAcc:{:.4f} PairwiseF1:{:.4f} TimePerStep:{:.4f}s TotalTiming:{}m{}s".format( - epoch, step_counter, loss_, softmaxlossterm, circlelossterm, acc_, f1, timingperstep, (timing) // 60, - timing % 60)) - - if (step_counter+1) % 500 == 0: - saver.save(sess, model_save_path+'.ckpt', global_step=step_counter+1) - tf.io.write_graph(sess.graph, model_save_path, 'graph.pbtxt', as_text=True) - print('save mode to {}'.format(model_save_path)) - - # if (i+1) % 10 == 0: - # t0=time.time() - # images, groundtruth_lists = get_next_batch(marketdata,train=False) - # test_dict = {x: images, - # y_: groundtruth_lists, - # is_training: True, - # lr:lr_new} - # summary,loss_val,acc_val = sess.run([merged,loss, accuracy], feed_dict=test_dict) - # test_writer.add_summary(summary, i) - # t1 = time.time() - # test_text = 'Step: {}, Loss: {:.4f}, Accuracy: {:.4f}, Timing:{:.4f}s'.format( - # i+1, loss_val, acc_val,t1-t0) - # print(test_text) - - train_writer.close() - diff --git a/TensorFlow/contrib/cv/CircleLoss_ID1221_for_TensorFlow/train_performance_1p.sh b/TensorFlow/contrib/cv/CircleLoss_ID1221_for_TensorFlow/train_performance_1p.sh new file mode 100644 index 0000000000000000000000000000000000000000..f9d478d61b6a896b05d3a1c9ed8748239d66cf84 --- /dev/null +++ b/TensorFlow/contrib/cv/CircleLoss_ID1221_for_TensorFlow/train_performance_1p.sh @@ -0,0 +1,4 @@ +python3 test_npu.py \ + --lfw_dir ./dataset/lfw-deepfunneled_align \ + --data_dir ./dataset/MS-Celeb-1M_clean_align \ + --pretrained_model ./20220602-013518/model-20220602-013518.ckpt-258 \ No newline at end of file diff --git a/TensorFlow/contrib/cv/CircleLoss_ID1221_for_TensorFlow/train_testcase.sh b/TensorFlow/contrib/cv/CircleLoss_ID1221_for_TensorFlow/train_testcase.sh deleted file mode 100644 index faf25281e347fe6c690603affdf0da2c4553742e..0000000000000000000000000000000000000000 --- a/TensorFlow/contrib/cv/CircleLoss_ID1221_for_TensorFlow/train_testcase.sh +++ /dev/null @@ -1,31 +0,0 @@ -#! /bin/bash -#开发者个人独立预置的数据集、预训练模型、ATC-OM模型等，支持从OBS仓下载 -obsutil cp obs://liuyixin2/datasets/Market-1501-v15.09.15/resnet_v1_50.ckpt ./dataset/ -f -r -obsutil cp obs://liuyixin2/datasets/Market-1501-v15.09.15/bounding_box_train.zip ./dataset/ -f -r -unzip -d ./dataset/ ./dataset/bounding_box_train.zip -pip3 install -r requirements.txt - - -#testcase主体，开发者根据不同模型写作 -batch_size=16 -epochs=1 -data_url="./dataset/" -train_url="./dataset/" - -nohup python3.7 train_npuv3.py \ - --data_url="${data_url}" \ - --epoch_size="${epochs}" \ - --train_url="${train_url}" > train.log 2>&1 - -wait - -#结果判断，功能检查输出ckpt/日志关键字、精度检查loss值/accucy关键字、性能检查耗时打点/ThroughOutput等关键字 -key1="Train success" #功能检查字 -key2="images/sec" #性能检查字 -key3="log10" #精度检查字 - -if [ `grep -c "$key1" "tran.log"` -ne '0' ] ;then #可以根据需要调整检查逻辑 - echo "Run testcase success!" -else - echo "Run testcase failed!" -fi \ No newline at end of file diff --git a/TensorFlow/contrib/cv/CircleLoss_ID1221_for_TensorFlow/trainmulitigpu.py b/TensorFlow/contrib/cv/CircleLoss_ID1221_for_TensorFlow/trainmulitigpu.py deleted file mode 100644 index cfe4c1219d7a5b6f866ed0f70e12796ea05932cf..0000000000000000000000000000000000000000 --- a/TensorFlow/contrib/cv/CircleLoss_ID1221_for_TensorFlow/trainmulitigpu.py +++ /dev/null @@ -1,247 +0,0 @@ -# Copyright 2017 The TensorFlow Authors. All Rights Reserved. -# -# Licensed under the Apache License, Version 2.0 (the "License"); -# you may not use this file except in compliance with the License. -# You may obtain a copy of the License at -# -# http://www.apache.org/licenses/LICENSE-2.0 -# -# Unless required by applicable law or agreed to in writing, software -# distributed under the License is distributed on an "AS IS" BASIS, -# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. -# See the License for the specific language governing permissions and -# limitations under the License. -# ============================================================================ -# Copyright 2021 Huawei Technologies Co., Ltd -# -# Licensed under the Apache License, Version 2.0 (the "License"); -# you may not use this file except in compliance with the License. -# You may obtain a copy of the License at -# -# http://www.apache.org/licenses/LICENSE-2.0 -# -# Unless required by applicable law or agreed to in writing, software -# distributed under the License is distributed on an "AS IS" BASIS, -# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. -# See the License for the specific language governing permissions and -# limitations under the License. - -import time -import numpy as np - -import tensorflow as tf -from tensorflow.contrib import slim -from tensorflow.examples.tutorials.mnist import input_data - - -from src.dataflow.market import MarketTriplet -import numpy as np -import skimage.transform -import tensorflow as tf -from config import * -from model import * -from data import * -from loss import * - -from tool import copy_code -copy_code(src="/home/liulizhao/projects/liuyixin/projects",tar=log_dir+"code/") - - -import os -gpu_ids = args.gpu -os.environ["CUDA_VISIBLE_DEVICES"]=gpu_ids - -# mnist = input_data.read_data_sets("mnist/", one_hot=True) -gpu_ids = gpu_ids.split(',') -gpu_ids = [int(id) for id in gpu_ids] -num_gpus = len(gpu_ids) -print("Available GPU Number :"+str(num_gpus)) - -# 读入数据 -dataset = tf_dataset(batch_size=batch_size*num_gpus,epoch_size=None) -iterator = dataset.make_one_shot_iterator() -one_element = iterator.get_next() - -def get_available_gpus(): - """ - code from http://stackoverflow.com/questions/38559755/how-to-get-current-available-gpus-in-tensorflow - """ - from tensorflow.python.client import device_lib as _device_lib - local_device_protos = _device_lib.list_local_devices() - return [x.name for x in local_device_protos if x.device_type == 'GPU'] - -def average_gradients(tower_grads): - average_grads = [] - for grad_and_vars in zip(*tower_grads): - grads = [] - for g, _ in grad_and_vars: - expend_g = tf.expand_dims(g, 0) - grads.append(expend_g) - grad = tf.concat(grads, 0) - grad = tf.reduce_mean(grad, 0) - v = grad_and_vars[0][1] - grad_and_var = (grad, v) - average_grads.append(grad_and_var) - return average_grads - -PS_OPS = ['Variable', 'VariableV2', 'AutoReloadVariable'] -def assign_to_device(device, ps_device='/cpu:0'): - def _assign(op): - node_def = op if isinstance(op, tf.NodeDef) else op.node_def - if node_def.op in PS_OPS: - return "/" + ps_device - else: - return device - - return _assign - -def ft_resnet50(x,istrain=False): - from tensorflow.contrib.slim import nets - keep_rate = 1.0 - is_training = False - if istrain: - keep_rate = args.droprate - is_training = True - with slim.arg_scope(nets.resnet_v1.resnet_arg_scope()): - net, endpoints = nets.resnet_v1.resnet_v1_50(x, num_classes=None, - is_training=is_training,global_pool=True) - import math - with tf.variable_scope('Logits'): - with slim.arg_scope([slim.conv2d, slim.fully_connected], - weights_regularizer=slim.l2_regularizer(5e-4)): - resnetfeat = tf.squeeze(net, axis=[1, 2]) - fc1 = slim.fully_connected(resnetfeat, num_outputs=num_bottleneck, - activation_fn=None,weights_initializer =tf.random_normal_initializer(mean=0,stddev=math.sqrt(2/num_bottleneck))) - bn1 = slim.batch_norm(fc1, activation_fn=None,is_training=is_training) - drop1 = slim.dropout(bn1, keep_prob=keep_rate) - norm_feat = tf.math.l2_normalize(drop1,axis=1,name='features',epsilon=1e-12) - logits = slim.fully_connected(drop1, num_outputs=num_classes, - activation_fn=None,weights_initializer=tf.random_normal_initializer(mean=0,stddev=0.001)) - return logits,norm_feat - - -def get_varlist(): - checkpoint_exclude_scopes = 'Logits' - exclusions = None - if checkpoint_exclude_scopes: - exclusions = [ - scope.strip() for scope in checkpoint_exclude_scopes.split(',')] - variables_to_restore = [] - add_var = [] - - for var in slim.get_model_variables(): - excluded = False - for exclusion in exclusions: - if var.op.name.startswith(exclusion): - excluded = True - if not excluded: - variables_to_restore.append(var) - else: - add_var.append(var) - return variables_to_restore,add_var - - -config = tf.ConfigProto(log_device_placement=True, allow_soft_placement = True) -config.gpu_options.allow_growth = True - - -def adjust_LR(step,epoch): - lr_l = args.init_lr * 0.1**((epoch)//40) - lr_b = 0.1 * lr_l - if epoch < warmup: - factor = min(1.0, 0.1 + 0.9 / (12936//(batch_size*num_gpus)*warmup)) - lr_l,lr_b = factor*lr_l,factor*lr_b - return lr_l,lr_b - -# def train(): -with tf.device("/cpu:0"): - global_step=tf.train.get_or_create_global_step() - tower_grads = [] - - x = tf.placeholder(tf.float32, [None, im_size[0],im_size[1],3],name='inputs') - y_ = tf.placeholder(tf.int32, [None],name='labels') - lr = tf.placeholder(tf.float32, name='lr') - is_training = tf.placeholder(tf.bool, name='is_training') - keep_rate = tf.placeholder(tf.float32, name='keep_rate') - - optimizer = tf.train.MomentumOptimizer(learning_rate=lr,momentum=0.9,use_nesterov=True) - # optimizer2 = tf.train.MomentumOptimizer(learning_rate=0.05,momentum=0.9,use_nesterov=True) - - with tf.variable_scope(tf.get_variable_scope()): - for i in gpu_ids: - with tf.device(assign_to_device('/gpu:{}'.format(i), ps_device='/cpu:0')): - _x = x[i * batch_size:(i + 1) * batch_size] - _y = y_[i * batch_size:(i + 1) * batch_size] - logits,norm_feat = ft_resnet50(_x,True) - - variables_to_restore,add_var = get_varlist() - - circle_loss_op,sp,sn,recall,precision,f1_score = PairWiseCircleLoss(norm_feat,_y) - softmaxloss = tf.nn.sparse_softmax_cross_entropy_with_logits( - labels=_y, logits=logits) - softmaxloss = tf.reduce_mean(softmaxloss) - circle_loss_term = lamda * circle_loss_op/batch_size - loss = softmaxloss + circle_loss_term - grads = optimizer.compute_gradients(loss) - # grads2 = optimizer2.compute_gradients(loss,var_list=add_var) - tower_grads.append(grads) - # tower_grads.append(grads2) - softmaxlogits = tf.nn.softmax(logits) - classes = tf.argmax(softmaxlogits, axis=1, name='classes') - accuracy_softmax = tf.reduce_mean(tf.cast( - tf.equal(tf.cast(classes, dtype=tf.int32), _y), dtype=tf.float32)) - - tf.summary.histogram('sn',sn) - tf.summary.histogram('sp',sp) - # recall,precision,f1_score - tf.summary.scalar('pairwise_recall',recall) - tf.summary.scalar('pairwise_precision',precision) - tf.summary.scalar('pairwise_f1_score',f1_score) - tf.summary.scalar('softmaxloss',softmaxloss) - # tf.summary.scalar('reg_loss',reg_term) - tf.summary.scalar('circleloss',circle_loss_term) - tf.summary.scalar('totalloss',loss) - tf.summary.scalar('accuracy_softmax',accuracy_softmax) - tf.summary.scalar('lr',lr) - merged = tf.summary.merge_all() - grads = average_gradients(tower_grads) - train_op = optimizer.apply_gradients(grads) - sample_counter=0 - with tf.Session(config=config) as sess: - - train_writer = tf.summary.FileWriter(log_dir + '/train', sess.graph) - - sess.run(tf.global_variables_initializer()) - # print(variables_to_restore) - saver_restore = tf.train.Saver(var_list=variables_to_restore) - saver_restore.restore(sess, resnet_model_path) - saver = tf.train.Saver(tf.global_variables()) - epoch = 0 - stepcounter = 0 - # for step in range(1, 50000): - t0 = time.time() - while epoch < epoch_size: - ts = time.time() - stepcounter+=1 - images, groundtruth_lists = sess.run(one_element) - batch_now = len(images) - sample_counter+=batch_now - epoch = sample_counter // 12936 - lr_l,lr_b = adjust_LR(stepcounter,epoch) - train_dict = {x: images, - y_: groundtruth_lists, - is_training: True, - lr:lr_l, - keep_rate:args.droprate} - summary,_ = sess.run([merged,train_op], feed_dict=train_dict) - train_writer.add_summary(summary, stepcounter) - te = time.time() - ts - if stepcounter % 10 == 0 or stepcounter == 1: - loss_value, acc,f1 = sess.run([loss, accuracy_softmax,f1_score], feed_dict=train_dict) - # print(f"Epoch:{epoch} Step:" + str(stepcounter) + " loss:" + str(loss_value) + " softmaxAcc:" + str(acc)+f' f1score:{f1}'+", %i sample/sec" % int(batch_now/te)) - timing = int(time.time()-t0) - print("Epoch:{} Step:{} TrainLoss:{:.4f} SoftmaxAcc:{:.4f} PairwiseF1:{:.4f} BatchRate:{}/sec Timing:{}m{}s".format(epoch,stepcounter,loss_value,acc,f1,int(batch_now/te),(timing)//60,timing%60)) - if (epoch+1) % 10 == 0: - saver.save(sess, model_save_path, global_step=stepcounter+1) - print('save mode to {}'.format(model_save_path)) -train_writer.close() \ No newline at end of file