MomentSeg: Moment-Centric Sampling for Enhanced Video Pixel Understanding

Benchmark Model Rank Results
temporal-sentence-grounding-on-activitynet-captions-temporal-localization-val-2MomentSeg-3B ([FIND] matching) (extra data)#2mIoU: 49.3R1@0.5: 52.0R1@0.7: 29.0
referring-video-object-segmentation-on-revosMomentSeg-7B#3J&F: 65.1J: 62.3F: 67.8
referring-video-object-segmentation-on-revosMomentSeg-3B#5J&F: 62.6J: 60.0F: 65.2
moment-retrieval-on-charades-staMomentSeg-3B (θ=0.3) (extra data)#51R@1 IoU=0.5: 55.9R@1 IoU=0.7: 25.3R@1 IoU=0.3: 78.9mIoU: 50.7