xERM.TDE.public

Cross-Domain Empirical Risk Minimization for Unbiased Long-tailed Classification

This project is built on top of TDE

Install

Please refer to TDE to set environment and dataset.

Places365-LT

Please change the resnet50 = load_model(resnet50, pretrain='/data1/pretrained/resnet50-0676ba61.pth') in models/ResNet50Feature.py to your resnet50 pretrained model path.

Please change the ResNet152 pretrained model path in models/ResNet152Feature.py

ResNet50

Step 1: Train TDE model

python main.py --cfg ./config/Places_LT/resnet50_TDE.yaml --gpu 0,1,2,3

Step 2: Train xERM-TDE model

python main.py --cfg ./config/Places_LT/resnet50_xERM.yaml --gpu 0,1,2,3 --xERM

Step 3: Evaluate

python main.py --cfg ./config/Places_LT/resnet50_xERM.yaml --gpu 0,1,2,3 --xERM --test

ResNet152

Step 1: Train TDE model

python main.py --cfg ./config/Places_LT/resnet152_TDE.yaml --gpu 0,1,2,3

Step 2: Train xERM-TDE model

python main.py --cfg ./config/Places_LT/resnet152_xERM.yaml --gpu 0,1,2,3 --xERM

Step 3: Evaluate

python main.py --cfg ./config/Places_LT/resnet152_xERM.yaml --gpu 0,1,2,3 --xERM --test

Citation

if you find our codes helpful, please cite our paper:

@inproceedings{beierxERM,
  title={Cross-Domain Empirical Risk Minimization for Unbiased Long-tailed Classification},
  author={Zhu, Beier and Niu, Yulei and Hua, Xian-Sheng and Zhang, Hanwang},
  booktitle={AAAI Conference on Artificial Intelligence},
  year={2022}
}