IMAGE SEGMENTATION WITH INCONSISTENCY MASKS

Summary from the TLDR AI newsletter

Inconsistency Masks (IM) are a new method for image segmentation that works even with limited data. This approach, tested on the ISIC 2018 dataset, beats traditional techniques and even outperforms models trained on fully labeled datasets.