Action Recognition on Kinetics 400 (test val)
87.4Top-1 AccVideoMAE320
Evaluation results
| Method | Links | ||
|---|---|---|---|
| VideoMAE320Backbone=ViT-H, Extra data=no external data, Ex. labels=false, Frames=32, GFLOPs=7397x4x3, Param=633, resolution=3202022.03 | 87.4 | 97.6 | |
| MaskFeat+352Backbone=MVIT-L, Extra data=Kinetics-600, Ex. labels=false, Frames=40, GFLOPs=3790x4x3, Param=2182022.03 | 87 | 97.4 | |
| VideoMAEBackbone=ViT-H, Extra data=no external data, Ex. labels=false, Frames=16, GFLOPs=1192x5x3, Param=6332022.03 | 86.6 | 97.1 | |
| VideoMAE320Backbone=ViT-L, Extra data=no external data, Ex. labels=false, Frames=32, GFLOPs=3958x4x3, Param=305, resolution=3202022.03 | 86.1 | 97.3 | |
| VideoMAEBackbone=ViT-L, Extra data=no external data, Ex. labels=false, Frames=16, GFLOPs=597x5x3, Param=3052022.03 | 85.2 | 96.8 | |
| ViViTBackbone=ViT-H, Extra data=JFT-300M, Ex. labels=true, Frames=32, GFLOPs=3981x4x32022.03 | 84.9 | 95.8 | |
| MaskFeatBackbone=MVIT-L, Extra data=no external data, Ex. labels=false, Frames=16, GFLOPs=377x10x1, Param=2182022.03 | 84.3 | 96.3 | |
| ViViT FEBackbone=ViT-L, Extra data=JFT-300M, Ex. labels=true, Frames=128, GFLOPs=3980x1x32022.03 | 83.5 | 94.3 | |
| Video SwinBackbone=Swin-L, Extra data=ImageNet-21K, Ex. labels=true, Frames=32, GFLOPs=604x4x3, Param=1972022.03 | 83.1 | 95.9 | |
| ViViT FEBackbone=ViT-L, Extra data=ImageNet-21K, Ex. labels=true, Frames=128, GFLOPs=3980x1x32022.03 | 81.7 | 93.8 | |
| VideoMAEBackbone=ViT-B, Extra data=no external data, Ex. labels=false, Frames=16, GFLOPs=180x5x3, Param=872022.03 | 81.5 | 95.1 | |
| TimeSformerBackbone=ViT-L, Extra data=ImageNet-21K, Ex. labels=true, Frames=96, GFLOPs=8353x1x3, Param=4302022.03 | 80.7 | 94.7 | |
| BEVTBackbone=Swin-B, Extra data=IN-1K+DALLE, Ex. labels=false, Frames=32, GFLOPs=282x4x3, Param=882022.03 | 80.6 | — | |
| MotionformerBackbone=ViT-L, Extra data=ImageNet-21K, Ex. labels=true, Frames=32, GFLOPs=1185x10x3, Param=3822022.03 | 80.2 | 94.8 | |
| MViTv1Backbone=MViTv1-B, Extra data=no external data, Ex. labels=false, Frames=32, GFLOPs=170x5x1, Param=372022.03 | 80.2 | 94.4 | |
| SlowFastBackbone=R101+NL, Extra data=no external data, Ex. labels=false, Frames=16+64, GFLOPs=234x10x3, Param=602022.03 | 79.8 | 93.9 | |
| TDN EnBackbone=ResNet101, Ex. labels=true, Frames=8+16, GFLOPs=198x10x3, Param=882022.03 | 79.4 | 94.4 | |
| TANetBackbone=ResNet152, Extra data=ImageNet-1K, Ex. labels=true, Frames=16, GFLOPs=242x4x3, Param=592022.03 | 79.3 | 94.1 | |
| VideoMAEBackbone=ViT-S, Extra data=no external data, Ex. labels=false, Frames=16, GFLOPs=57x5x3, Param=222022.03 | 79 | 93.8 | |
| ip-CSNBackbone=ResNet152, Extra data=no external data, Ex. labels=false, Frames=32, GFLOPs=109x10x3, Param=332022.03 | 77.8 | 92.8 | |
| VIMPACBackbone=ViT-L, Extra data=HowTo100M+DALLE, Ex. labels=false, Frames=10, GFLOPs=N/Ax10x3, Param=3072022.03 | 77.4 | — | |
| NL I3DBackbone=ResNet101, Ex. labels=true, Frames=128, GFLOPs=359x10x3, Param=622022.03 | 77.3 | 93.3 |