Video Classification on Kinetics-400 (Corruption Robustness)
78.4Average Accuracy (Corruption Robustness)DINOv2
Evaluation results
| Method | Links | |||||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| DINOv2Arch=ViT-g/14, Protocol=Frozen linear probe, Input frames=82023.04 | 78.4 | — | — | — | — | — | — | — | — | — | — | — | — | |
| OpenCLIPArch=ViT-G/14, Protocol=Frozen linear probe, Input frames=82023.04 | 78.3 | — | — | — | — | — | — | — | — | — | — | — | — | |
| DINOv2Arch=ViT-L/14, Protocol=Frozen linear probe, Input frames=82023.04 | 76.3 | — | — | — | — | — | — | — | — | — | — | — | — | |
| DINOv2Arch=ViT-B/14, Protocol=Frozen linear probe, Input frames=82023.04 | 73.2 | — | — | — | — | — | — | — | — | — | — | — | — | |
| iBOTArch=ViT-L/16, Protocol=Frozen linear probe, Input frames=82023.04 | 72.6 | — | — | — | — | — | — | — | — | — | — | — | — | |
| DINOv2Arch=ViT-S/14, Protocol=Frozen linear probe, Input frames=82023.04 | 67.8 | — | — | — | — | — | — | — | — | — | — | — | — | |
| DINOArch=ViT-B/8, Protocol=Frozen linear probe, Input frames=82023.04 | 64.5 | — | — | — | — | — | — | — | — | — | — | — | — | |
| MAEArch=ViT-H/14, Protocol=Frozen linear probe, Input frames=82023.04 | 54.2 | — | — | — | — | — | — | — | — | — | — | — | — | |
| VITTABackbone=TANet, Batch Size=82022.11 | 48.94 | 47.95 | 43.9 | 34.73 | 64.07 | 50.08 | 49.23 | 52.8 | 57.69 | 63.89 | 46.77 | 61.88 | 14.24 | |
| TENTBackbone=TANet, Batch Size=82022.11 | 41.13 | 34.16 | 31.54 | 22.64 | 60.29 | 46.03 | 34.88 | 45.54 | 52.5 | 61.87 | 32.5 | 59.21 | 12.44 | |
| NORMBackbone=TANet, Batch Size=82022.11 | 39.32 | 32.55 | 30.09 | 21.67 | 58.22 | 43.44 | 33.25 | 43.99 | 50.74 | 60.28 | 28.19 | 57.14 | 12.29 | |
| T3ABackbone=TANet, Batch Size=82022.11 | 37.59 | 29.38 | 27.19 | 16.24 | 58.02 | 40.43 | 30.33 | 46.5 | 53.05 | 60.39 | 26.5 | 49.16 | 13.93 | |
| SHOTBackbone=TANet, Batch Size=82022.11 | 37.43 | 32.24 | 29.91 | 21.76 | 55.72 | 40.26 | 32.95 | 42.77 | 47.93 | 56.44 | 23.59 | 51.05 | 14.55 | |
| Source-OnlyBackbone=TANet, Batch Size=82022.11 | 37.16 | 28.52 | 26.14 | 15.97 | 57.89 | 39.66 | 29.51 | 47.01 | 52.79 | 60.6 | 25.69 | 48.23 | 13.91 | |
| DUABackbone=TANet, Batch Size=12022.11 | 31.88 | 24.26 | 22.25 | 14.68 | 50.74 | 35.1 | 25.04 | 38.15 | 43.41 | 53.29 | 18.32 | 48.04 | 9.31 |