Multi-Object Tracking on TAO 1.0 (val)
36.3Base TETARegionCLIP + OVTrack
Evaluation Results
| Method | Links | ||||||||
|---|---|---|---|---|---|---|---|---|---|
| RegionCLIP + OVTrackClasses (Base)=false, Classes (Novel)=true, Data (CC3M)=true, Data (LVIS)=true, Data (TAO)=false, Backbone=ResNet-502023.04 | 36.3 | 53.9 | 36.3 | 18.7 | 32 | 51.4 | 33.2 | 11.4 | |
| OVTrackClasses (Base)=true, Classes (Novel)=false, Data (CC3M)=false, Data (LVIS)=true, Data (TAO)=false, Backbone=ResNet-502023.04 | 35.5 | 49.3 | 36.9 | 20.2 | 27.8 | 48.8 | 33.6 | 1.5 | |
| MASA (Grounding-DINO features)Supervision Type=Self-supervised, Domain Setting=Zero-shot, Association Features=Grounding-DINO2024.06 | 34.9 | 51.8 | 37.6 | 15.4 | — | — | — | — | |
| MASA (Detic features)Supervision Type=Self-supervised, Domain Setting=Zero-shot, Association Features=Detic2024.06 | 34.7 | 51.9 | 36.4 | 15.8 | — | — | — | — | |
| TETerSupervision Type=Fully-supervised, Domain Setting=In-domain, Object Detection=Same observations as ours2024.06 | 34.6 | 52.1 | 36.7 | 15 | — | — | — | — | |
| MASA (SAM-B features)Supervision Type=Self-supervised, Domain Setting=Zero-shot, Association Features=SAM-B2024.06 | 34.5 | 51.9 | 36.6 | 15.1 | — | — | — | — | |
| MASA (SAM-H features)Supervision Type=Self-supervised, Domain Setting=Zero-shot, Association Features=SAM-H2024.06 | 34.5 | 51.8 | 36.4 | 15.4 | — | — | — | — | |
| MASA (R50 features)Supervision Type=Self-supervised, Domain Setting=Zero-shot, Association Features=ResNet-502024.06 | 34.1 | 52.1 | 35.7 | 15 | — | — | — | — | |
| TETerClasses (Base)=true, Classes (Novel)=true, Data (CC3M)=false, Data (LVIS)=true, Data (TAO)=true, Backbone=ResNet-502023.04 | 30.3 | 47.4 | 31.6 | 12.1 | 25.7 | 45.9 | 31.1 | 0.2 | |
| RegionCLIP + Tracktor++Classes (Base)=false, Classes (Novel)=true, Data (CC3M)=true, Data (LVIS)=true, Data (TAO)=false, Backbone=ResNet-502023.04 | 29.6 | 52.4 | 19.6 | 16.9 | 25.7 | 50.1 | 18.9 | 8.1 | |
| RegionCLIP + DeepSORTClasses (Base)=false, Classes (Novel)=true, Data (CC3M)=true, Data (LVIS)=true, Data (TAO)=false, Backbone=ResNet-502023.04 | 28.4 | 52.5 | 15.6 | 17 | 24.5 | 49.2 | 15.3 | 9 | |
| Tracktor++ (ViLD)Classes (Base)=true, Classes (Novel)=false, Data (CC3M)=false, Data (LVIS)=true, Data (TAO)=false, Backbone=ResNet-502023.04 | 28.3 | 47.4 | 20.5 | 17 | 22.7 | 46.7 | 19.3 | 2.2 | |
| QDTrackClasses (Base)=true, Classes (Novel)=false, Data (CC3M)=false, Data (LVIS)=true, Data (TAO)=true, Backbone=ResNet-502023.04 | 27.1 | 45.6 | 24.7 | 11 | 22.5 | 42.7 | 24.4 | 0.4 | |
| DeepSORT (ViLD)Classes (Base)=true, Classes (Novel)=false, Data (CC3M)=false, Data (LVIS)=true, Data (TAO)=false, Backbone=ResNet-502023.04 | 26.9 | 47.1 | 15.8 | 17.7 | 21.1 | 46.4 | 14.7 | 2.3 |