diff --git "a/docs/models/\344\272\272\350\204\270\346\243\200\346\265\213\345\244\215\347\216\260\346\265\201\347\250\213.md" "b/docs/models/\344\272\272\350\204\270\346\243\200\346\265\213\345\244\215\347\216\260\346\265\201\347\250\213.md"
new file mode 100644
index 0000000000000000000000000000000000000000..b5f7b432f02f9ff694a0a01ccdc9e5d7d8fccaff
--- /dev/null
+++ "b/docs/models/\344\272\272\350\204\270\346\243\200\346\265\213\345\244\215\347\216\260\346\265\201\347\250\213.md"
@@ -0,0 +1,237 @@
+# 1 安装系统环境（ubuntu）
+## 1.1 安装显卡驱动
+
+- 打开software & updates ->选择**Additional Drivers** 安装一个驱动
+- 使用 nvidia-smi 查看显卡是否正常
+## 1.2 安装CUDA
+1.2.1 下载CUDA
+ - i https://developer.nvidia.com/cuda-toolkit-archive
+ - 选择对应版本的CUDA
+ - 选择 -> Linux x86_64 -> Ubuntu-> version -> Runfile (local)
+**下面这个是根据版本选择得到的**
+wget https://developer.download.nvidia.com/compute/cuda/11.8.0/local_installers/cuda_11.8.0_520.61.05_linux.run
+sudo sh cuda_11.8.0_520.61.05_linux.run
+等待一段时间，会弹出一个窗口，选择continue 回车 输入 **accept**->进入CUDA Installer界面 选择Driver 点击**空格**取消选择Driver 
+--> Install
+安装完成后，需要进行环境变量配置
+输入 vim ~/.bashrc
+添加以下内容：
+ - export PATH=/usr/local/cuda-11.8/bin${PATH:+:${PATH}}
+ - export LD_LIBRARY_PATH=/usr/local/cuda-11.8/lib64\${LD_LIBRARY_PATH:+:${LD_LIBRARY_PATH}}
+
+ - 输入**nvcc -V** 查看是否安装成功
+
+## 1.3 安装cudnn
+1.3.1 下载cudnn
+点击下面的链接，进行对应版本的选择
+ - https://developer.nvidia.com/rdp/cudnn-archive
+ - sudo apt-get install 
+
+## 1.4 安装 Nvidia-docker
+进入官网
+https://docs.nvidia.com/datacenter/cloud-native/container-toolkit/latest/install-guide.html
+选择 install with apt
+ - Configure the production repository:
+  curl -fsSL https://nvidia.github.io/libnvidia-container/gpgkey | sudo gpg --dearmor -o /usr/share/keyrings/nvidia-container-toolkit-keyring.gpg \
+  && curl -s -L https://nvidia.github.io/libnvidia-container/stable/deb/nvidia-container-toolkit.list | \
+    sed 's#deb https://#deb [signed-by=/usr/share/keyrings/nvidia-container-toolkit-keyring.gpg] https://#g' | \
+    sudo tee /etc/apt/sources.list.d/nvidia-container-toolkit.list
+ - Update the packages list from the repository:
+    sudo apt-get update
+ - Install the NVIDIA Container Toolkit packages:
+    sudo apt-get install -y nvidia-container-toolkit
+
+# 2 配置训练环境
+## 2.1 拉取docker镜像
+- modelscope github 仓库地址
+- https://github.com/modelscope/modelscope
+
+GPU镜像
+# py37
+- registry.cn-hangzhou.aliyuncs.com/modelscope-repo/modelscope:ubuntu20.04-cuda11.3.0-py37-torch1.11.0-tf1.15.5-1.6.1
+
+# py38
+- registry.cn-hangzhou.aliyuncs.com/modelscope-repo/modelscope:ubuntu20.04-cuda11.8.0-py38-torch2.0.1-tf2.13.0-1.9.5
+
+我拉取的是 py38 版本
+- docker pull registry.cn-hangzhou.aliyuncs.com/modelscope-repo/modelscope:ubuntu20.04-cuda11.8.0-py38-torch2.0.1-tf2.13.0-1.9.5
+查看镜像
+- docker images
+启动镜像 -v /home/ubuntu/:/home/ubuntu/ 本地文件映射到容器内
+- docker run -it --gpus all -v /home/ubuntu/:/home/ubuntu/ registry.cn-hangzhou.aliyuncs.com/modelscope-repo/modelscope:ubuntu20.04-cuda11.8.0-py38-torch2.0.1-tf2.13.0-1.9.5 -n /bin/bash 
+
+## 2.2 拉取训练模型
+- git clone https://github.com/biubug6/Pytorch_Retinaface.git
+
+
+# 3 数据集制作
+## 3.1 数据集标注
+- 数据集标注使用LabelMe进行数据标注
+文件夹样式
+- Dataset
+  - annotations
+  - images
+  - flags.txt
+- 目标数据格式
+```markdown
+  ./data/widerface/
+    train/
+      images/
+      label.txt
+    val/
+      images/
+      wider_val.txt
+```
+- 标注内容 
+  - 人脸边界框：上额头到下额头，左右脸边缘
+  - 左眼、右眼、鼻子、左嘴角、右嘴角
+- 标签文件转换
+```python
+import os
+import json
+import random  # 导入random模块
+
+def convert_json_to_format(json_data, folder_name):
+    # 提取图片路径
+    image_path = json_data['imagePath']
+
+    # 获取图像尺寸
+    image_width = json_data['imageWidth']
+    image_height = json_data['imageHeight']
+
+    # 初始化输出字符串
+    output = ""
+
+    # 初始化各部位坐标值为 -1
+    face_box_x, face_box_y, face_box_w, face_box_h = -1, -1, -1, -1
+    left_eye_x, left_eye_y = -1, -1
+    right_eye_x, right_eye_y = -1, -1
+    nose_x, nose_y = -1, -1
+    left_mouse_x, left_mouse_y = -1, -1
+    right_mouse_x, right_mouse_y = -1, -1
+
+    # 处理形状数据
+    shapes = json_data['shapes']
+
+    for shape in shapes:
+        label = shape['label']
+        points = shape['points']
+
+        if label == 'face' and len(points) == 2:
+            # 转换矩形框坐标
+            x1, y1 = points[0]
+            x2, y2 = points[1]
+            face_box_x = min(x1, x2)
+            face_box_y = min(y1, y2)
+            face_box_w = abs(x2 - x1)
+            face_box_h = abs(y2 - y1)
+        elif label == 'left_eye':
+            # 添加左眼位置信息
+            left_eye_x, left_eye_y = points[0]
+        elif label == 'right_eye':
+            # 添加右眼位置信息
+            right_eye_x, right_eye_y = points[0]
+        elif label == 'nose':
+            # 添加鼻子位置信息
+            nose_x, nose_y = points[0]
+        elif label == 'left_mouse':
+            # 添加左嘴角位置信息
+            left_mouse_x, left_mouse_y = points[0]
+        elif label == 'right_mouse':
+            # 添加右嘴角位置信息
+            right_mouse_x, right_mouse_y = points[0]
+
+    # 构建输出字符串
+    output = f"{int(face_box_x)} {int(face_box_y)} {int(face_box_w)} {int(face_box_h)} "
+    output += f"{float(left_eye_x):.3f} {float(left_eye_y):.3f} 0.0 "
+    output += f"{float(right_eye_x):.3f} {float(right_eye_y):.3f} 0.0 "
+    output += f"{float(nose_x):.3f} {float(nose_y):.3f} 0.0 "
+    output += f"{float(left_mouse_x):.3f} {float(left_mouse_y):.3f} 0.0 "
+    output += f"{float(right_mouse_x):.3f} {float(right_mouse_y):.3f} 0.0 "
+
+    # 添加随机置信度值，范围在0.8到0.99之间
+    confidence = random.uniform(0.8, 0.99)
+    output += f"{confidence:.2f}"
+
+    # 返回结果
+    return f"# {folder_name}/{os.path.basename(image_path)}\n{output}\n"
+
+def process_folder(folder_path, output_file):
+    # 打开输出文件
+    with open(output_file, 'w', encoding='utf-8') as out_file:
+        # 遍历文件夹中的所有文件
+        for filename in os.listdir(folder_path):
+            if filename.endswith('.json'):
+                file_path = os.path.join(folder_path, filename)
+                with open(file_path, 'r', encoding='utf-8') as file:
+                    json_data = json.load(file)
+                    # 获取文件夹名
+                    folder_name = os.path.basename(os.path.dirname(file_path))
+                    result = convert_json_to_format(json_data, folder_name)
+                    out_file.write(result)
+
+# 指定文件夹路径和输出文件路径
+folder_path = r'C:\Users\zhb20\Desktop\FaceData\train\images\face'
+output_file = r'C:\Users\zhb20\Desktop\FaceData\train\label.txt'
+
+# 调用函数
+process_folder(folder_path, output_file)
+```
+验证集标签文件制作
+```python
+import os
+
+def save_folder_and_image_names(folder_path, output_file):
+    """
+    遍历指定文件夹中的所有图片，将文件夹名和图片名组合后保存到输出文件中。
+
+    :param folder_path: 包含图片的文件夹路径
+    :param output_file: 输出文件的路径
+    """
+    # 检查文件夹是否存在
+    if not os.path.isdir(folder_path):
+        print(f"错误：{folder_path} 不是一个有效的文件夹路径")
+        return
+
+    # 图片文件可能的扩展名列表
+    image_extensions = ['.jpg', '.jpeg', '.png', '.gif', '.bmp', '.tiff']
+
+    with open(output_file, 'w', encoding='utf-8') as file:
+        for root, dirs, files in os.walk(folder_path):
+            for name in files:
+                # 获取文件扩展名
+                ext = os.path.splitext(name)[1].lower()
+                if ext in image_extensions:
+                    # 组合文件夹名和图片名
+                    folder_name = os.path.basename(root)
+                    full_name = f"{folder_name}/{name}"
+                    file.write(full_name + '\n')
+
+if __name__ == "__main__":
+    # 修改这里的文件夹路径和输出文件路径
+    folder_path = r'C:\Users\Administrator\Desktop\facedata\val\images\face'
+    output_file = r'C:\Users\Administrator\Desktop\facedata\val\wider_val.txt'
+
+    save_folder_and_image_names(folder_path, output_file)
+    print(f"图片信息已保存至 {output_file}")
+```
+# 4 windows 环境搭建
+## 4.1 conda安装
+进入网址
+- https://anaconda.org/anaconda/conda
+    下载conda 默认步骤安装即可
+- conda create -n 环境名 python==版本号 -y
+- conda activate 环境名
+- 配置国内镜像源地址
+- pip install -i https://mirrors.tuna.tsinghua.edu.cn/pypi/web/simple some-package
+- 安装 pytorch 2.0.1
+- [pytorch下载网址](https://pytorch.org/get-started/locally/)
+- pillow==8.2.0
+- opencv-python==4.5.2.54
+- numpy==1.22.0
+- matplotlib==3.5.1
+- tensorboard==2.5.0
+-大概就可以使用了，如果出现版本不兼容，灵活调整
+**注**
+- 如果要使用GPU训练，请先安装环境
+- 显卡驱动、CUDA、CUDNN
\ No newline at end of file
diff --git a/example/vision/face_detection/README.md b/example/vision/face_detection/README.md
index 784ae33bd4fb8bb94b1cf924d092ce644e8cc0c1..427bd4694661a06bd07bda52b2627b0fbcdc28c8 100644
--- a/example/vision/face_detection/README.md
+++ b/example/vision/face_detection/README.md
@@ -33,7 +33,7 @@
 
 ## 2 运行前的准备
 
-* 请确保你已经下载了 [凌智视觉模块人脸检测模型](https://gitee.com/LockzhinerAI/LockzhinerVisionModule/releases/download/v0.0.0/LZ-RetinaFace.rknn)
+* 请确保你已经下载了 [凌智视觉模块人脸检测模型](https://gitee.com/LockzhinerAI/LockzhinerVisionModule/releases/download/v0.0.3/LZ-RetinaFace-2024-1121-1627.rknn)
 
 ## 3 在凌智视觉模块上部署模型
 
diff --git a/example/vision/face_detection/images/img-1.png b/example/vision/face_detection/images/img-1.png
new file mode 100644
index 0000000000000000000000000000000000000000..f224ef199470251b43ef28220c2fd1a7a7448e04
Binary files /dev/null and b/example/vision/face_detection/images/img-1.png differ
diff --git a/example/vision/face_detection/python/README.md b/example/vision/face_detection/python/README.md
index 6495c8d14473e8db3ad10b9637a51eaccc372c66..1264cd36b67b1f018afc29bcdc55e0e1b4a7c955 100644
--- a/example/vision/face_detection/python/README.md
+++ b/example/vision/face_detection/python/README.md
@@ -278,7 +278,10 @@ python test_retina_face.py LZ-RetinaFace.rknn
 
 连接凌智视觉模块图片传输助手[凌智视觉模块图片传输助手下载地址](https://gitee.com/LockzhinerAI/LockzhinerVisionModule/releases/download/v0.0.0/LockzhinerVisionModuleImageFetcher.exe)后，选择连接设备
 
-![](../images/img.png)
+![](../images/img-1.png)
+
+
+
 
 运行程序后，屏幕上开始打印矩形框信息和置信度，并在一段时间后输出 FPS 值
 
diff --git a/example/vision/face_detection/python/test_retina_face_latest.py b/example/vision/face_detection/python/test_retina_face_latest.py
new file mode 100644
index 0000000000000000000000000000000000000000..a8865d4766d4febd3eedcbad6dd15d467e830661
--- /dev/null
+++ b/example/vision/face_detection/python/test_retina_face_latest.py
@@ -0,0 +1,78 @@
+from lockzhiner_vision_module.cv2 import imread, imwrite, VideoCapture
+from lockzhiner_vision_module.vision import RetinaFace, visualize
+from lockzhiner_vision_module.edit import Edit
+import time
+import sys
+
+
+def predict_video(face_det_model, width, height):
+    edit = Edit()
+    edit.start_and_accept_connection()
+
+    video_capture = VideoCapture()
+    video_capture.set_width(width)
+    video_capture.set_height(height)
+    if video_capture.open(0) is False:
+        print("Failed to open capture")
+        exit(1)
+
+    while True:
+        read_index = 0
+        total_time_ms = 0
+        for i in range(30):
+            ret, mat = video_capture.read()
+            if ret is False:
+                continue
+
+            start_time = time.time()
+            results = face_det_model.predict(mat)
+            end_time = time.time()
+            total_time_ms += end_time - start_time
+            read_index += 1
+            print(f"result size is {len(results)}")
+            for result in results:
+                box = result.box
+                score = result.score
+                print(
+                    f"(x,y,w,h,score): [{box.x},{box.y},{box.width},{box.height},{score}]"
+                )
+            vis_mat = visualize(mat, results)
+            edit.print(vis_mat)
+        print(f"FPS is {1.0 / (total_time_ms/read_index)}")
+
+
+def predict_image(face_det_model, image_path):
+    image = imread(image_path)
+    results = face_det_model.predict(image)
+    vis_mat = visualize(image, results)
+    print(f"result size is {len(results)}")
+    for result in results:
+        box = result.box
+        score = result.score
+        print(
+            f"(x,y,w,h,score): [{box.x},{box.y},{box.width},{box.height},{score}]"
+        )
+    imwrite("face_det.png", vis_mat)
+    
+
+if __name__ == "__main__":
+    args = sys.argv
+    if len(args) < 2:
+        print("Need model path. Example: python test_retina_face.py LZ-RetinaFace.rknn width height")
+        exit(1)
+
+    model = RetinaFace()
+    if model.initialize(args[1]) is False:
+        print("Failed to initialize RetinaFace")
+        exit(1)
+
+    video_width = 640
+    video_height = 480
+    if len(args) == 5:
+        video_width = int(args[3])
+        video_height = int(args[4])
+
+    if len(args) == 2:
+        predict_video(model, video_width, video_height)
+    elif len(args) == 3:
+        predict_image(model, args[2])
\ No newline at end of file