Long-term Point Tracking on TAP-Vid DAVIS 480p (test)
66.4Avg Temporal ErrorTAP-Net
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| TAP-NetMode=S (Supervised), Backbone=CNN2023.04 | 66.4 | 79 | 46 | |
| RAFTBackbone=CNN2023.04 | 66.7 | — | — | |
| DINOv2Backbone=ViT-L-14/72023.04 | 66.7 | — | — | |
| DINOv2 LoRA-2LMode=TT (Test-time Training), Param.=0.07M, Backbone=ViT-L-14/72023.04 | 73.2 | 84.8 | 58 | |
| PIPs++Mode=S (Supervised), Backbone=CNN2023.04 | 73.6 | — | — | |
| OmnimotionMode=TT (Test-time Training), Backbone=MLP2023.04 | 74.1 | 84.2 | 58.4 | |
| TAPIRMode=S (Supervised), Backbone=CNN2023.04 | 77.3 | 89.5 | 65.7 | |
| DropDINO LoRA-2LMode=TT (Test-time Training), Param.=0.05M, Backbone=ViT-B-16/82023.04 | 79 | 89.1 | 64.9 | |
| Co-TrackerMode=S (Supervised), Backbone=Transf.2023.04 | 79.4 | 89.5 | 65.6 | |
| DropDINO LoRA-4LMode=TT (Test-time Training), Param.=0.10M, Backbone=ViT-B-16/82023.04 | 79.5 | 89.6 | 65.4 | |
| DropDINO LoRA-6LMode=TT (Test-time Training), Param.=0.15M, Backbone=ViT-B-16/82023.04 | 79.7 | 89.8 | 65.7 | |
| DINO-TrackerMode=TT (Test-time Training), Param.=7.59M, Backbone=CNN2023.04 | 80.4 | 88.1 | 64.6 |