Video Action Recognition on Epic-Kitchens 100 (test)
46.1Top-1 Action AccuracyORViT MF-HR-Light
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| ORViT MF-HR-LightPretrain=IN21K+K400, Params (10^6)=1142021.10 | 46.1 | 68.5 | 58.7 | |
| ORViT MF-HRPretrain=IN21K+K400, Params (10^6)=1482021.10 | 45.7 | 68.4 | 58.7 | |
| Mformer-HRPre-train=IN-21K+K-4002021.06 | 44.5 | 67 | 58.5 | |
| MF-HRPretrain=IN-21K+K400, Params (10^6)=1092021.10 | 44.5 | 67 | 58.5 | |
| MF-HR + STINPretrain=IN-21K+K400, Params (10^6)=1112021.10 | 44.2 | 67 | 57.9 | |
| Mformer-LPre-train=IN-21K+K-4002021.06 | 44.1 | 67.1 | 57.6 | |
| MF-LPretrain=IN-21K+K400, Params (10^6)=1092021.10 | 44.1 | 67.1 | 57.6 | |
| MF-HR + STRG + STINPretrain=IN-21K+K400, Params (10^6)=1192021.10 | 44.1 | 66.9 | 57.8 | |
| ViViT-LPre-train=IN-21K+K-4002021.06 | 44 | 66.4 | 56.8 | |
| ViViT-LPretrain=IN-21K+K4002021.10 | 44 | 66.4 | 56.8 | |
| MformerPre-train=IN-21K+K-4002021.06 | 43.1 | 66.7 | 56.5 | |
| MFPretrain=IN-21K+K400, Params (10^6)=1092021.10 | 43.1 | 66.7 | 56.5 | |
| MF-HR + STRGPretrain=IN-21K+K400, Params (10^6)=1172021.10 | 42.5 | 65.8 | 55.4 | |
| SlowFastPre-train=K-4002021.06 | 38.5 | 65.6 | 50 | |
| SlowFastPretrain=K4002021.10 | 38.5 | 65.6 | 50 | |
| TSMPre-train=IN-1K2021.06 | 38.3 | 67.9 | 49 | |
| TSMPretrain=IN-1K2021.10 | 38.3 | 67.9 | 49 | |
| TBNPre-train=IN-1K2021.06 | 36.7 | 66 | 47.2 | |
| TBNPretrain=IN-1K2021.10 | 36.7 | 66 | 47.2 | |
| TRNPre-train=IN-1K2021.06 | 35.3 | 65.9 | 45.4 | |
| TRNPretrain=IN-1K2021.10 | 35.3 | 65.9 | 45.4 | |
| TSNPre-train=IN-1K2021.06 | 33.2 | 60.2 | 46 | |
| TSNPretrain=IN-1K2021.10 | 33.2 | 60.2 | 46 | |
| TimeSformerPretrain=IN-21K, Params (10^6)=1212021.10 | 32.9 | 55.8 | 50.1 |