End-to-End Temporal Action Detection with 1B Parameters Across 1000 Frames

Benchmark Model Rank Results
temporal-action-localization-on-activitynetAdaTAD (VideoMAEv2-giant)#4mAP: 41.93mAP IOU@0.5: 61.72mAP IOU@0.75: 43.35mAP IOU@0.95: 10.85
temporal-action-localization-on-epic-kitchensAdaTAD (verb, VideoMAE-L)#1Avg mAP (0.1-0.5): 29.3mAP IOU@0.1: 33.1mAP IOU@0.2: 32.2
temporal-action-localization-on-thumos14AdaTAD (VideoMAEv2-giant)#1Avg mAP (0.3:0.7): 76.9mAP IOU@0.3: 89.7mAP IOU@0.4: 86.7