Awesome

Semantic-Aware Domain Generalized Segmentation

This is the code related to "Semantic-Aware Domain Generalized Segmentation" (CVPR 2022)

1. Paper

Semantic-Aware Domain Generalized Segmentation
IEEE/CVF Computer Vision and Pattern Recognition (CVPR 2022)

If you find it helpful to your research, please cite as follows:

@inproceedings{peng2022semantic,
  title={Semantic-Aware Domain Generalized Segmentation},
  author={Peng, Duo and Lei, Yinjie and Hayat, Munawar and Guo, Yulan and Li, Wen},
  booktitle={Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR)},
  year={2022},
  publisher={IEEE}
}

2. Preparation

• PyTorch 1.7.1
• CUDA 11.1
• Python 3.7.9
• Torchvision 0.8.2
• Download the model pretrained on ImageNet. Put it into each file named as (pretianed_model).

3. Datasets

• Download GTA5 datasets.
• Download SYNTHIA.
• Download Cityscapes.
• Download BDDS.
• Download Mapillary.

4. Usage

You can directly training the network within the following python file by using IDE such as Pycharm.

tools\train.py

We present the demo on GTAV to Cityscapes, other Settings can be implemented by converting the following args. (Except for the GTAV training set we use 640×640 resolution, all Settings are consistent with RobustNet, including dataloader, dataset partitioning etc. For more information, please refer to RobustNet).

arg_parser.add_argument('--data_root_path', type=str, default='XXX')  # Training data path

arg_parser.add_argument('--list_path', type=str, default='XXX')  # Training data list path

arg_parser.add_argument('--dataset', type=str, default='XXX')  # Training dataset

arg_parser.add_argument('--val_dataset', type=str, default='XXX')  # Validation/Testing dataset

We present the demo on ResNet50. You can change the backbone by converting the following args.

arg_parser.add_argument('--backbone', type=str, default='XXX')  # Network Backbone

5. Results

update status

The code (V1) is uploaded. (2022-06-19)