DomainDrop: Suppressing Domain-Sensitive Channels for Domain Generalization

Requirements

Python == 3.7.3
Pytorch == 1.8.1
Cuda == 10.1
Torchvision == 0.4.2
Tensorflow == 1.14.0
GPU == RTX 2080Ti

DataSets

Please download PACS dataset from here. Make sure you use the official train/val/test split in PACS paper. Take /data/DataSets/ as the saved directory for example:

images -> /data/DataSets/PACS/kfold/art_painting/dog/pic_001.jpg, ...
splits -> /data/DataSets/PACS/pacs_label/art_painting_crossval_kfold.txt, ...

Then set the "data_root" as "/data/DataSets/" and "data" as "PACS" in both train_domain.py and train.sh.

Training

For training the model, please set the "result_path" where the results are saved in both train_domain.py and train.sh. Then simply running the code to train a ResNet-18:

python train_domain.py --target [domain_index] --device [GPU_index]

The domain_index denotes the index of target domain, and GPU_index denotes the GPU device number.

domain_index: [0:'photo', 1:'art_painting', 2:'cartoon', 3:'sketch']

Or run the train.sh directly.

Evaluation

To evaluate the performance of the models, you can download the models trained on PACS as below:

Target domain	Photo	Art	Cartoon	Sketch
Acc(%)	96.71	84.91	80.72	84.32
models	download	download	download	download

Please set the --eval = 1 and --eval_model_path as the saved path of the downloaded models. e.g., /trained/model/path/photo/model.pt. Then you can simple run:

python train_domain.py --target [domain_index] --device [GPU_index] --eval 1 --eval_model_path '/trained/model/path/photo/model.pt'

Citations

@inproceedings{guo2023domaindrop,
  title={DomainDrop: Suppressing Domain-Sensitive Channels for Domain Generalization},
  author={Guo, Jintao and Qi, Lei and Shi, Yinghuan},
  booktitle={Proceedings of the IEEE/CVF International Conference on Computer Vision},
  year={2023}
}

Acknowledgement

Part of our code is derived from the following repository.

MMLD: "Domain Generalization Using a Mixture of Multiple Latent Domains", AAAI 2020

We thank to the authors for releasing their codes. Please also consider citing their work.