Action recognition on EPFL-Smart-Kitchen-30
46.6Overall AccuracyLLaVAction-7B
Evaluation results
| Method | Links | |||
|---|---|---|---|---|
| LLaVAction-7BEvaluation Protocol=Trained model2025.03 | 46.6 | 49.7 | 27 | |
| Multi-modal VideoMAEEvaluation Protocol=Trained model2025.03 | 40 | 43.6 | 19.4 | |
| VideoMAEEvaluation Protocol=Trained model2025.03 | 37.5 | 41.1 | 16.6 | |
| LLaVAction-7BEvaluation Protocol=Zero-shot generalization of the second-stage model2025.03 | 36.2 | 38.1 | 24.6 | |
| LLaVA-Video-7BEvaluation Protocol=Zero-shot generalization of the second-stage model2025.03 | 22.5 | 22.9 | 18.8 | |
| AVIONEvaluation Protocol=Zero-shot generalization of the second-stage model2025.03 | 19.3 | 21.2 | 8.6 |