Keystep recognition on CrossTask
64.5AccuracyOurs
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| OursDownstream architecture=Transformer2023.07 | 64.5 | — | |
| DistantSup.Downstream architecture=Transformer2023.07 | 64.2 | — | |
| TimeSformerDownstream architecture=Transformer2023.07 | 60.9 | — | |
| VideoCLIPDownstream architecture=Transformer2023.07 | 60.1 | — | |
| OursDownstream architecture=MLP2023.07 | 59.5 | — | |
| DistantSup.Downstream architecture=MLP2023.07 | 58.4 | — | |
| TimeSformerDownstream architecture=MLP2023.07 | 55.5 | — | |
| VideoCLIPDownstream architecture=MLP2023.07 | 51.3 | — | |
| SlowFastDownstream architecture=Transformer2023.07 | 48.5 | — | |
| S3DDownstream architecture=Transformer2023.07 | 45.3 | — | |
| SlowFastDownstream architecture=MLP2023.07 | 44.9 | — | |
| S3DDownstream architecture=MLP2023.07 | 39.9 | — | |
| Ours2023.07 | 30.5 | 0.16 | |
| Graph2Vid2023.07 | 24.8 | 0.168 | |
| Drop-DTW2023.07 | 22.3 | 0.151 | |
| Bag of Steps2023.07 | 20.5 | 0.137 | |
| Random2023.07 | 11 | 0.047 |