Knowing What, Where and When to Look: Efficient Video Action Modeling with Attention

Benchmark Model Rank Results
action-recognition-in-videos-on-somethingTSM+W3 (16 frames, RGB ResNet-50)Top-1 Accuracy: 66.5Top-5 Accuracy: 90.4
action-recognition-in-videos-on-something-1TSM+W3 (16 frames, ResNet50)Top 1 Accuracy: 52.6Top 5 Accuracy: 81.3