VIRST: Video-Instructed Reasoning Assistant for SpatioTemporal Segmentation

Benchmark Model Rank Results
referring-video-object-segmentation-on-mevisVIRST#2J&F: 62.9J: 60.4F: 65.4
referring-video-object-segmentation-on-revosVIRST#2J&F: 68.4J: 66.3F: 70.6
referring-video-object-segmentation-on-ref-davis17VIRST#4J&F: 79.5