UnLoc: A Unified Framework for Video Localization Tasks

Benchmark Model Rank Results
action-segmentation-on-coinUnLoc-L#1Frame accuracy: 72.8
moment-retrieval-on-charades-staUnLoc-L#11R@1 IoU=0.5: 60.8R@1 IoU=0.7: 38.4R@5 IoU=0.5: 88.2
moment-retrieval-on-charades-staUnLoc-B#17R@1 IoU=0.5: 58.1R@1 IoU=0.7: 35.4R@5 IoU=0.5: 87.4
moment-retrieval-on-qvhighlightsUnLoc-L#29R@1 IoU=0.5: 66.1R@1 IoU=0.7: 46.7
moment-retrieval-on-qvhighlightsUnLoc-B#30R@1 IoU=0.5: 64.5R@1 IoU=0.7: 48.8
natural-language-moment-retrieval-onUnLoc-L#4R@1,IoU=0.5: 48.3R@1,IoU=0.7: 30.2R@5,IoU=0.5: 79.2
natural-language-moment-retrieval-onUnLoc-B#5R@1,IoU=0.5: 48.0R@1,IoU=0.7: 29.7R@5,IoU=0.5: 81.5
temporal-action-localization-on-activitynetUnLoc-L#29mAP IOU@0.5: 59.3