arxivcs.CV2026-07-08
`Attention-Guided Cross-Temporal Clustering for Self-Supervised Video Object Segmentation
Waqas Arshid, Mohammad Awrangjeb, Alan Wee-Chung Liew, Yongsheng Gao
Video object segmentation (VOS) is a fundamental task in video understanding, requiring accurate delineation and consistent tracking of objects across frames. While supervised methods achieve strong performance, they rely on densely annotated datasets that are costly to obtain an…