Spatiotemporal Self-attention Modeling with Temporal Patch Shift for Action Recognition

Benchmark Model Rank Results
action-classification-on-kinetics-400TPS#69Acc@1: 82.5
action-recognition-in-videos-on-somethingTPS#37Top-1 Accuracy: 69.8
action-recognition-in-videos-on-something-1TPS#10Top 1 Accuracy: 58.3
action-recognition-on-diving-48PSB#7Accuracy: 86