Space-time Reinforcement Network for Video Object Segmentation

arXiv:2405.04042v1 Announce Type: new
Abstract: Recently, video object segmentation (VOS) networks typically use memory-based methods: for each query frame, the mask is predicted by space-time matching to memory frames. Despite these methods having superior performance, they suffer from two issues: 1) Challenging data can destroy the space-time coherence between adjacent video frames. 2) Pixel-level matching will lead to undesired mismatching caused by the noises or distractors. To address the aforementioned issues, we first propose to generate an auxiliary frame between …

abstract arxiv cs.ai cs.cv data memory network networks object performance query reinforcement segmentation space type video