RSANet-R50 (8+16 frames, ImageNet pretrained, 2 clips

Reported on 4 benchmarks across 2 tasks · 1 paper

Note: results are matched by exact model name. Different papers may use the same name for different model variants.

Robots2 results

Activity RecognitiononSomething-Something V2
Top-1 Accuracy· uses extra data· 2021-11-02
67.7
best: 77.3 (MVD (Kinetics400 pretrain, ViT-H, 16 frame))
Relational Self-Attention: What's Missing in Attention for Video Understanding arXiv:2111.01673
Activity RecognitiononSomething-Something V2
Top-5 Accuracy· uses extra data· 2021-11-02
91.1
best: 96.3 (DejaVid)
Relational Self-Attention: What's Missing in Attention for Video Understanding arXiv:2111.01673

Time Series2 results

Action RecognitiononSomething-Something V2
Top-1 Accuracy· uses extra data· 2021-11-02
67.7
best: 77.3 (MVD (Kinetics400 pretrain, ViT-H, 16 frame))
Relational Self-Attention: What's Missing in Attention for Video Understanding arXiv:2111.01673
Action RecognitiononSomething-Something V2
Top-5 Accuracy· uses extra data· 2021-11-02
91.1
best: 96.3 (DejaVid)
Relational Self-Attention: What's Missing in Attention for Video Understanding arXiv:2111.01673