Weakly Supervised Incremental Segmentation via Semantic Anchors and Spatial Arbitration

Arxiv CS.CV (computer vision)

Introduces a weakly supervised incremental segmentation method using semantic anchors and spatial arbitration to address noisy supervision in continual learning. Aims to reduce feature drift and semantic corruption.

Arxiv CS.CV (computer vision)·Zhonggai WangKai FangGuangyu Gao

7 days ago