Video Classification on Mini-Kinetics-200 (val)
81.8Top-1 AccDSA+I3D
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| DSA+I3DBackbone=3D ResNet50, Ttrain x Utrain=4 x 4, Tinfer x Uinfer x #crop=4 x 8 x 32021.05 | 81.8 | 95.4 | |
| V4D+I3DBackbone=3D ResNet50, Ttrain x Utrain=4 x 4, Tinfer x Uinfer x #crop=4 x 10 x 32021.05 | 80.7 | 95.3 | |
| I3D+CGNLBackbone=3D ResNet101, Ttrain x Utrain=32 x 1, Tinfer x Uinfer x #crop=32 x 10 x 32021.05 | 79.9 | 93.4 | |
| I3D+NLBackbone=3D ResNet101, Ttrain x Utrain=32 x 1, Tinfer x Uinfer x #crop=32 x 10 x 32021.05 | 79.2 | 93.2 | |
| S3DBackbone=S3D Inception, Ttrain x Utrain=64 x 1, Tinfer x Uinfer x #crop=N/A2021.05 | 78.9 | — | |
| I3D+CGNLBackbone=3D ResNet50, Ttrain x Utrain=32 x 1, Tinfer x Uinfer x #crop=32 x 10 x 32021.05 | 78.8 | 94.4 | |
| I3D+NLBackbone=3D ResNet50, Ttrain x Utrain=32 x 1, Tinfer x Uinfer x #crop=32 x 10 x 32021.05 | 77.5 | 94 | |
| I3DBackbone=3D ResNet101, Ttrain x Utrain=32 x 1, Tinfer x Uinfer x #crop=32 x 10 x 32021.05 | 77.4 | 93.2 | |
| DSA+I3DBackbone=3D ResNet18, Ttrain x Utrain=4 x 4, Tinfer x Uinfer x #crop=4 x 8 x 32021.05 | 77.3 | 93.9 | |
| V4D+I3DBackbone=3D ResNet18, Ttrain x Utrain=4 x 4, Tinfer x Uinfer x #crop=4 x 10 x 32021.05 | 75.6 | 92.7 | |
| I3DBackbone=3D ResNet50, Ttrain x Utrain=32 x 1, Tinfer x Uinfer x #crop=32 x 10 x 32021.05 | 75.5 | 92.2 |