Video Recognition on Kinetics-400 (val)
81.2AccuracyMViTv2-B
Evaluation Results
| Method | Links | |
|---|---|---|
| MViTv2-BPretrain=IN-1K, Type=Transformer, Params=51.2M2023.11 | 81.2 | |
| TimeSFormerPretrain=K400, Type=Transformer, Params=122M2023.11 | 80.7 | |
| SlowFast-101Pretrain=IN-1K, Type=ConvNet+RNN, Params=62.8M2023.11 | 79.8 | |
| UniRepLKNet-SPretrain=-, Type=ConvNet, Params=55.5M2023.11 | 54.8 | |
| ImageBindPretrain=CLIP Data, Type=Transformer, Params=632M2023.11 | 50 | |
| Meta-TransformerPretrain=LAION-2B, Type=Transformer, Params=86.9M2023.11 | 47.3 |