Interactive Spatiotemporal Token Attention Network for Skeleton-based General Interactive Action Recognition

Benchmark Model Rank Results
3d-action-recognition-on-assembly101ISTA-Net#5Actions Top-1: 28.07Verbs Top-1: 62.66Object Top-1: 31.69
action-recognition-on-h2o-2-hands-and-objectsISTA-Net#5Actions Top-1: 89.09RGB: NoHand Pose: 3DObject Pose: Yes
human-interaction-recognition-on-ntu-rgbd-120ISTA-Net#3Accuracy (Cross-Setup): 91.7Accuracy (Cross-Subject): 90.5
skeleton-based-action-recognition-on-h2o-2ISTA-Net#4Accuracy: 89.09±1.21