Flexible Sampling for Long-tailed Skin Lesion Classification

by Lie Ju, Yicheng Wu, Lin Wang, Zhen Yu, Xin Zhao, Xin Wang, Paul Bonnington and Zongyuan Ge*

Intro.

This is an official Pytorch implementation of our paper published in MICCAI 2022.

Flexible Sampling for Long-tailed Skin Lesion Classification

The code is avaliable now!

We have also provided other widely-used tricks for long-tailed learning. Feel free to use them!

Usage

Please follow:

1. On the 8-class dataset, please download ISIC 2019 dataset.

2. Then you can use our provided split training, validation, and test data, all stored in a NumPy format. Please use 'np.load()' to extract the information.

3. For the 14-class dataset, please use official API to download the extra images from the ISIC dataset gallery using our provided .csv file.

4. Also, the split data is provided. Please use 'np.load()' to extract the information.

Citation

If you find this repository is helpful for your work, please cite our work:

@inproceedings{ju2022flexible,
  title={Flexible Sampling for Long-Tailed Skin Lesion Classification},
  author={Ju, Lie and Wu, Yicheng and Wang, Lin and Yu, Zhen and Zhao, Xin and Wang, Xin and Bonnington, Paul and Ge, Zongyuan},
  booktitle={Medical Image Computing and Computer Assisted Intervention--MICCAI 2022: 25th International Conference, Singapore, September 18--22, 2022, Proceedings, Part III},
  pages={462--471},
  year={2022},
  organization={Springer}
}

Reference

[1] Cui, Yin, et al. "Class-balanced loss based on effective number of samples." Proceedings of the IEEE/CVF conference on computer vision and pattern recognition. 2019.

[2] Lin, Tsung-Yi, et al. "Focal loss for dense object detection." Proceedings of the IEEE international conference on computer vision. 2017.

[3] Zhang, Songyang, et al. "Distribution alignment: A unified framework for long-tail visual recognition." Proceedings of the IEEE/CVF conference on computer vision and pattern recognition. 2021.

[4] Cao, Kaidi, et al. "Learning imbalanced datasets with label-distribution-aware margin loss." Advances in neural information processing systems 32 (2019).

[5] Zhang, Hongyi, et al. "mixup: Beyond empirical risk minimization." arXiv preprint arXiv:1710.09412 (2017).