Skip to content

Latest commit

 

History

History
 
 

image_classification

README.md

Domain Generalization for Image Classification

Installation

Example scripts can deal with WILDS datasets. You should first install wilds before using these scripts.

pip install wilds

Example scripts also support all models in PyTorch-Image-Models. You also need to install timm to use PyTorch-Image-Models.

pip install timm

Dataset

Following datasets can be downloaded automatically:

Supported Methods

Experiment and Results

The shell files give the script to reproduce the benchmarks with specified hyper-parameters. For example, if you want to reproduce IRM on Office-Home, use the following script

# Train with IRM on Office-Home Ar Cl Rw -> Pr task using ResNet 50.
# Assume you have put the datasets under the path `data/office-home`, 
# or you are glad to download the datasets automatically from the Internet to this path
CUDA_VISIBLE_DEVICES=0 python irm.py data/office-home -d OfficeHome -s Ar Cl Rw -t Pr -a resnet50 --seed 0 --log logs/irm/OfficeHome_Pr

For more information please refer to Get Started for help.

Citation

If you use these methods in your research, please consider citing.

@inproceedings{IBN-Net,  
    author = {Xingang Pan, Ping Luo, Jianping Shi, and Xiaoou Tang},  
    title = {Two at Once: Enhancing Learning and Generalization Capacities via IBN-Net},  
    booktitle = {ECCV},  
    year = {2018}  
}

@inproceedings{mixstyle,
    title={Domain Generalization with MixStyle},
    author={Zhou, Kaiyang and Yang, Yongxin and Qiao, Yu and Xiang, Tao},
    booktitle={ICLR},
    year={2021}
}

@inproceedings{MLDG,
    title={Learning to Generalize: Meta-Learning for Domain Generalization},
    author={Li, Da and Yang, Yongxin and Song, Yi-Zhe and Hospedales, Timothy},
    booktitle={AAAI},
    year={2018}
}
 
@misc{IRM,
    title={Invariant Risk Minimization}, 
    author={Martin Arjovsky and Léon Bottou and Ishaan Gulrajani and David Lopez-Paz},
    year={2020},
    eprint={1907.02893},
    archivePrefix={arXiv},
    primaryClass={stat.ML}
}

@inproceedings{VREx,
    title={Out-of-Distribution Generalization via Risk Extrapolation (REx)}, 
    author={David Krueger and Ethan Caballero and Joern-Henrik Jacobsen and Amy Zhang and Jonathan Binas and Dinghuai Zhang and Remi Le Priol and Aaron Courville},
    year={2021},
    booktitle={ICML},
}

@inproceedings{GroupDRO,
    title={Distributionally Robust Neural Networks for Group Shifts: On the Importance of Regularization for Worst-Case Generalization}, 
    author={Shiori Sagawa and Pang Wei Koh and Tatsunori B. Hashimoto and Percy Liang},
    year={2020},
    booktitle={ICLR}
}

@inproceedings{deep_coral,
    title={Deep coral: Correlation alignment for deep domain adaptation},
    author={Sun, Baochen and Saenko, Kate},
    booktitle={ECCV},
    year={2016},
}