DSCL

Pytorch implementation of Paper "Decoupled Contrastive Learning for Long-Tailed Recognition" (AAAI 2024)

Installation

1. Clone code

    git clone https://github.com/SY-Xuan/DSCL.git
    cd ./DSCL

2. Install Package

conda create -n pink python=3.10 -y
conda activate pink
pip install --upgrade pip # enable PEP 660 support
pip install -e .

Train

1. Representation Learning

You need to download ImageNet and change the data path in the train.sh. We use 2 RTX 3090.

sh ./train.sh

2. Classifier Learning

You need to change the checkpoint path in the train_cls.sh. We use 2 RTX 3090.

sh ./train_cls.sh

Citations

If you find this code useful for your research, please cite our paper:

@inproceedings{xuan2024decoupled,
  title={Decoupled Contrastive Learning for Long-Tailed Recognition},
  author={Xuan, Shiyu and Zhang, Shiliang},
  booktitle={Proceedings of the AAAI Conference on Artificial Intelligence},
  volume={38},
  number={6},
  pages={6396--6403},
  year={2024}
}

Contact me

If you have any questions about this code or paper, feel free to contact me at shiyu_xuan@stu.pku.edu.cn.

Acknowledgement

Codes are built upon moco and targeted-supcon. Thanks for these outstanding implementations.