Thank you for maintaining the list of papers on imbalanced learning!
Our paper: AUCSeg: AUC-oriented Pixel-level Long-tail Semantic Segmentation[NeurIPS 2024] extends imbalanced classification from the image level to the pixel level by introducing an AUC-based loss.
Hope it can be added to 2.5.4 Loss function engineering.
PDF
Code
Thank you once again for your efforts!
Best regards,