Video Action Recognition on Kinetics-700
82.9Top-1 AccEVA
Evaluation results
| Method | Links | ||
|---|---|---|---|
| EVAprotocol=fine-tuning2022.11 | 82.9 | — | |
| EVAPre-training Data=K-722 (merged K-400, K-600, K-700)2022.11 | 82.9 | — | |
| CoCaEvaluation Protocol=finetuned2022.05 | 82.7 | — | |
| CoCaEvaluation Protocol=finetuned2022.11 | 82.7 | — | |
| CoCaEvaluation Protocol=frozen2022.05 | 81.1 | — | |
| CoCaEvaluation Protocol=frozen2022.11 | 81.1 | — | |
| MaskFeat2022.05 | 80.4 | — | |
| MaskFeatprotocol=fine-tuning2022.11 | 80.4 | — | |
| MaskFeat2022.11 | 80.4 | — | |
| mPLUG-2Setting=Fine-tuning2023.02 | 80.4 | 94.9 | |
| MViTv2-L↑ 312^2, 40×3pre-train=IN-21K, FLOPs x views=2828×3×3, Param=217.62021.12 | 79.4 | 94.9 | |
| MoViNet2022.05 | 79.4 | — | |
| MoViNetSetting=Fine-tuning2023.02 | 79.4 | — | |
| CoVeR2022.05 | 78.5 | — | |
| CoVeR2022.11 | 78.5 | — | |
| CoVeRSetting=Fine-tuning2023.02 | 78.5 | 94.2 | |
| MViTv2-B, 32×3pre-train=scratch, FLOPs x views=206×3×3, Param=51.42021.12 | 76.6 | 93.2 | |
| mPLUG-2BaseSetting=Fine-tuning2023.02 | 74.6 | 91.2 | |
| MoViNet-A6pre-train=N/A, FLOPs x views=386×1×1, Param=31.42021.12 | 72.3 | — | |
| SlowFast 16×8+NLpre-train=K600, FLOPs x views=234×3×10, Param=59.92021.12 | 71 | 89.6 |