# DeepLabV3Plus-Pytorch **Repository Path**: cheng4632/DeepLabV3Plus-Pytorch ## Basic Information - **Project Name**: DeepLabV3Plus-Pytorch - **Description**: 同步 https://github.com/VainF/DeepLabV3Plus-Pytorch - **Primary Language**: Unknown - **License**: MIT - **Default Branch**: master - **Homepage**: None - **GVP Project**: No ## Statistics - **Stars**: 0 - **Forks**: 1 - **Created**: 2022-09-06 - **Last Updated**: 2022-09-06 ## Categories & Tags **Categories**: Uncategorized **Tags**: None ## README # DeepLabv3Plus-Pytorch DeepLabv3, DeepLabv3+ with pretrained models for Pascal VOC & Cityscapes. ## Quick Start ### 1. Available Architectures Specify the model architecture with '--model ARCH_NAME' and set the output stride using '--output_stride OUTPUT_STRIDE'. | DeepLabV3 | DeepLabV3+ | | :---: | :---: | |deeplabv3_resnet50|deeplabv3plus_resnet50| |deeplabv3_resnet101|deeplabv3plus_resnet101| |deeplabv3_mobilenet|deeplabv3plus_mobilenet | All pretrained models: [Dropbox](https://www.dropbox.com/sh/w3z9z8lqpi8b2w7/AAB0vkl4F5vy6HdIhmRCTKHSa?dl=0), [Tencent Weiyun](https://share.weiyun.com/qqx78Pv5) ### 2. Load the pretrained model: ```python model.load_state_dict( torch.load( CKPT_PATH )['model_state'] ) ``` ### 3. Visualize segmentation outputs: ```python outputs = model(images) preds = outputs.max(1)[1].detach().cpu().numpy() colorized_preds = val_dst.decode_target(preds).astype('uint8') # To RGB images, (N, H, W, 3), ranged 0~255, numpy array # Do whatever you like here with the colorized segmentation maps colorized_preds = Image.fromarray(colorized_preds[0]) # to PIL Image ``` ### 4. Atrous Separable Convolution **Note**: pre-trained models in this repo **do not** use Seperable Conv. Atrous Separable Convolution is supported in this repo. We provide a simple tool ``network.convert_to_separable_conv`` to convert ``nn.Conv2d`` to ``AtrousSeparableConvolution``. **Please run main.py with '--separable_conv' if it is required**. See 'main.py' and 'network/_deeplab.py' for more details. ### 5. Prediction Single image: ```bash python predict.py --input ~/Datasets/Cityscapes/leftImg8bit/train/bremen/bremen_000000_000019_leftImg8bit.png --dataset cityscapes --model deeplabv3plus_mobilenet --ckpt checkpoints/best_deeplabv3plus_mobilenet_cityscapes_os16.pth --save_val_results_to test_results ``` Image folder: ```bash python predict.py --input ~/Datasets/Cityscapes/leftImg8bit/train/bremen --dataset cityscapes --model deeplabv3plus_mobilenet --ckpt checkpoints/best_deeplabv3plus_mobilenet_cityscapes_os16.pth --save_val_results_to test_results ``` ## Results ### 1. Performance on Pascal VOC2012 Aug (21 classes, 513 x 513) Training: 513x513 random crop validation: 513x513 center crop | Model | Batch Size | FLOPs | train/val OS | mIoU | Dropbox | Tencent Weiyun | | :-------- | :-------------: | :----: | :-----------: | :--------: | :--------: | :----: | | DeepLabV3-MobileNet | 16 | 6.0G | 16/16 | 0.701 | [Download](https://www.dropbox.com/s/uhksxwfcim3nkpo/best_deeplabv3_mobilenet_voc_os16.pth?dl=0) | [Download](https://share.weiyun.com/A4ubD1DD) | | DeepLabV3-ResNet50 | 16 | 51.4G | 16/16 | 0.769 | [Download](https://www.dropbox.com/s/3eag5ojccwiexkq/best_deeplabv3_resnet50_voc_os16.pth?dl=0) | [Download](https://share.weiyun.com/33eLjnVL) | | DeepLabV3-ResNet101 | 16 | 72.1G | 16/16 | 0.773 | [Download](https://www.dropbox.com/s/vtenndnsrnh4068/best_deeplabv3_resnet101_voc_os16.pth?dl=0) | [Download](https://share.weiyun.com/iCkzATAw) | | DeepLabV3Plus-MobileNet | 16 | 17.0G | 16/16 | 0.711 | [Download](https://www.dropbox.com/s/0idrhwz6opaj7q4/best_deeplabv3plus_mobilenet_voc_os16.pth?dl=0) | [Download](https://share.weiyun.com/djX6MDwM) | | DeepLabV3Plus-ResNet50 | 16 | 62.7G | 16/16 | 0.772 | [Download](https://www.dropbox.com/s/dgxyd3jkyz24voa/best_deeplabv3plus_resnet50_voc_os16.pth?dl=0) | [Download](https://share.weiyun.com/uTM4i2jG) | | DeepLabV3Plus-ResNet101 | 16 | 83.4G | 16/16 | 0.783 | [Download](https://www.dropbox.com/s/bm3hxe7wmakaqc5/best_deeplabv3plus_resnet101_voc_os16.pth?dl=0) | [Download](https://share.weiyun.com/UNPZr3dk) | ### 2. Performance on Cityscapes (19 classes, 1024 x 2048) Training: 768x768 random crop validation: 1024x2048 | Model | Batch Size | FLOPs | train/val OS | mIoU | Dropbox | Tencent Weiyun | | :-------- | :-------------: | :----: | :-----------: | :--------: | :--------: | :----: | | DeepLabV3Plus-MobileNet | 16 | 135G | 16/16 | 0.721 | [Download](https://www.dropbox.com/s/753ojyvsh3vdjol/best_deeplabv3plus_mobilenet_cityscapes_os16.pth?dl=0) | [Download](https://share.weiyun.com/aSKjdpbL) | DeepLabV3Plus-ResNet101 | 16 | N/A | 16/16 | 0.762 | [Download](https://drive.google.com/file/d/1t7TC8mxQaFECt4jutdq_NMnWxdm6B-Nb/view?usp=sharing) | [Comming Soon]() #### Segmentation Results on Pascal VOC2012 (DeepLabv3Plus-MobileNet)
#### Segmentation Results on Cityscapes (DeepLabv3Plus-MobileNet)
#### Visualization of training ![trainvis](samples/visdom-screenshoot.png) ## Pascal VOC ### 1. Requirements ```bash pip install -r requirements.txt ``` ### 2. Prepare Datasets #### 2.1 Standard Pascal VOC You can run train.py with "--download" option to download and extract pascal voc dataset. The defaut path is './datasets/data': ``` /datasets /data /VOCdevkit /VOC2012 /SegmentationClass /JPEGImages ... ... /VOCtrainval_11-May-2012.tar ... ``` #### 2.2 Pascal VOC trainaug (Recommended!!) See chapter 4 of [2] The original dataset contains 1464 (train), 1449 (val), and 1456 (test) pixel-level annotated images. We augment the dataset by the extra annotations provided by [76], resulting in 10582 (trainaug) training images. The performance is measured in terms of pixel intersection-over-union averaged across the 21 classes (mIOU). *./datasets/data/train_aug.txt* includes the file names of 10582 trainaug images (val images are excluded). Please to download their labels from [Dropbox](https://www.dropbox.com/s/oeu149j8qtbs1x0/SegmentationClassAug.zip?dl=0) or [Tencent Weiyun](https://share.weiyun.com/5NmJ6Rk). Those labels come from [DrSleep's repo](https://github.com/DrSleep/tensorflow-deeplab-resnet). Extract trainaug labels (SegmentationClassAug) to the VOC2012 directory. ``` /datasets /data /VOCdevkit /VOC2012 /SegmentationClass /SegmentationClassAug # <= the trainaug labels /JPEGImages ... ... /VOCtrainval_11-May-2012.tar ... ``` ### 3. Training on Pascal VOC2012 Aug #### 3.1 Visualize training (Optional) Start visdom sever for visualization. Please remove '--enable_vis' if visualization is not needed. ```bash # Run visdom server on port 28333 visdom -port 28333 ``` #### 3.2 Training with OS=16 Run main.py with *"--year 2012_aug"* to train your model on Pascal VOC2012 Aug. You can also parallel your training on 4 GPUs with '--gpu_id 0,1,2,3' **Note: There is no SyncBN in this repo, so training with *multple GPUs and small batch size* may degrades the performance. See [PyTorch-Encoding](https://hangzhang.org/PyTorch-Encoding/tutorials/syncbn.html) for more details about SyncBN** ```bash python main.py --model deeplabv3plus_mobilenet --enable_vis --vis_port 28333 --gpu_id 0 --year 2012_aug --crop_val --lr 0.01 --crop_size 513 --batch_size 16 --output_stride 16 ``` #### 3.3 Continue training Run main.py with '--continue_training' to restore the state_dict of optimizer and scheduler from YOUR_CKPT. ```bash python main.py ... --ckpt YOUR_CKPT --continue_training ``` #### 3.4. Testing Results will be saved at ./results. ```bash python main.py --model deeplabv3plus_mobilenet --enable_vis --vis_port 28333 --gpu_id 0 --year 2012_aug --crop_val --lr 0.01 --crop_size 513 --batch_size 16 --output_stride 16 --ckpt checkpoints/best_deeplabv3plus_mobilenet_voc_os16.pth --test_only --save_val_results ``` ## Cityscapes ### 1. Download cityscapes and extract it to 'datasets/data/cityscapes' ``` /datasets /data /cityscapes /gtFine /leftImg8bit ``` ### 2. Train your model on Cityscapes ```bash python main.py --model deeplabv3plus_mobilenet --dataset cityscapes --enable_vis --vis_port 28333 --gpu_id 0 --lr 0.1 --crop_size 768 --batch_size 16 --output_stride 16 --data_root ./datasets/data/cityscapes ``` ## Reference [1] [Rethinking Atrous Convolution for Semantic Image Segmentation](https://arxiv.org/abs/1706.05587) [2] [Encoder-Decoder with Atrous Separable Convolution for Semantic Image Segmentation](https://arxiv.org/abs/1802.02611)