Spatio-temporal action localization on AVA v2.2 (val)
24.1mAP (Det)ConST-CL
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| ConST-CLBackbone=R3D50, Pre-training Dataset=K4002021.12 | 24.1 | 31.1 | |
| pBYOLBackbone=Slow-only, Pre-training Dataset=K4002021.12 | 23.4 | — | |
| VFS-inflatedBackbone=R3D50, Pre-training Dataset=K4002021.12 | 22.4 | 29.1 | |
| K400 supBackbone=R3D50, Pre-training Dataset=K4002021.12 | 22.2 | — | |
| pMoCoBackbone=Slow-only, Pre-training Dataset=K4002021.12 | 20.3 | — | |
| INet-supBackbone=I3D, Pre-training Dataset=K4002021.12 | 19.1 | 27.5 | |
| CVRLBackbone=R3D50, Pre-training Dataset=K4002021.12 | 18.4 | 24.9 |