Action Recognition on Moments in Time V1 (val)
48.7Top-1 AccUMT-L
Evaluation results
| Method | Links | ||
|---|---|---|---|
| UMT-LInput Size=8×384^2, FLOPs (G)=2440×12, Param (M)=3042023.03 | 48.7 | 78.2 | |
| UMT-LInput Size=8×224^2, FLOPs (G)=596×12, Param (M)=3042023.03 | 48 | 77.8 | |
| UniFormerV2-LInput Size=8×336^2, FLOPs (G)=1568×12, Param (M)=354, Extra data=web-scale data2023.03 | 47.8 | 76.9 | |
| MTV-HInput Size=32×280^2, FLOPs (G)=6130×12, Param (M)=1000+, Extra data=web-scale data2023.03 | 47.2 | 75.7 | |
| UniFormerV2-LInput Size=8×224^2, FLOPs (G)=666×12, Param (M)=354, Extra data=web-scale data2023.03 | 47 | 76.1 | |
| CoVeRInput Size=16×448^2, FLOPs (G)=5860×3, Param (M)=431, Extra data=web-scale data2023.03 | 46.1 | — | |
| MTV-HInput Size=32×224^2, FLOPs (G)=3706×12, Param (M)=1000+2023.03 | 45.6 | 74.7 | |
| UMT-BInput Size=8×384^2, FLOPs (G)=786×12, Param (M)=872023.03 | 45.5 | 74.6 | |
| UMT-BInput Size=8×224^2, FLOPs (G)=180×12, Param (M)=872023.03 | 44.6 | 74 | |
| UniFormerV2-BInput Size=8×224^2, FLOPs (G)=148×12, Param (M)=115, Extra data=web-scale data2023.03 | 42.7 | 71.5 | |
| ViViT-LInput Size=32×224^2, FLOPs (G)=3980×3, Param (M)=6122023.03 | 38.5 | 64.1 |