- Article
Video instance segmentation (VIS) is plagued by the high cost of pixel-level annotation and defects of weakly supervised segmentation, leading to the urgent need for a trade-off between annotation cost and performance. We propose a novel In-Depth Col...