For training the model, you may modify the run.sh
file. Define --data_folder
as the path to your dataset. We also provide our splits in the third section. To run, use bash run.sh
For testing the model, you may modify the run.sh
file. Set --resume
to True and define --ckpt
as the path to your save model. To run, use bash run.sh
We provide trained models with AdUni and different sampling technics with one train-test splits.
method | Dataset | Test acc. | Test f1 | url |
---|---|---|---|---|
AdUni | ISIC2018 | 87.1% | 79.0% | model |
AdUni+upsample | ISIC2018 | 88.2% | 79.8% | model |
AdUni | APTOS2019 | 83.3% | 69.4% | model |
AdUni+upsample | APTOS2019 | 83.9% | 71.1% | model |
@inproceedings{cong2022aduni, title={Adaptive Unified Contrastive Learning for Imbalanced Classification}, author={Cong Cong, Yixing Yang, Sidong Liu, Maurice Pagnucco, Antonio DiIeva, Shlomo Berkovsky, Yang Song}, booktitle={International Workshop on Machine Learning in Medical Imaging}, year={2022}, organization={Springer} }