IBoxCLA: Towards Robust Box-supervised Segmentation of Polyp via Improved Box-dice and Contrastive Latent-anchors

arXiv:2310.07248v3 Announce Type: replace
Abstract: Box-supervised polyp segmentation attracts increasing attention for its cost-effective potential. Existing solutions often rely on learning-free methods or pretrained models to laboriously generate pseudo masks, triggering Dice constraint subsequently. In this paper, we found that a model guided by the simplest box-filled masks can accurately predict polyp locations/sizes, but suffers from shape collapsing. In response, we propose two innovative learning fashions, Improved Box-dice (IBox) and Contrastive Latent-Anchors (CLA), and combine them to train a robust …

abstract anchors arxiv attention box cost cs.cv dice found free generate masks paper potential pretrained models replace robust segmentation solutions type via