Open-vocabulary video instance segmentation on SA-Co/VEval SmartGlasses (val)
53.9cgF1Human
Evaluation Results
| Method | Links | |||||||
|---|---|---|---|---|---|---|---|---|
| Human2025.11 | 53.9 | 91 | 59.6 | 68.1 | 53 | 87.9 | 71 | |
| SAM 32025.11 | 33.5 | 76 | 44.4 | 60.2 | 46.2 | 79.3 | 65 | |
| SAM 3temporal disambiguation=false2025.11 | 30.9 | 71 | 43.5 | 58.1 | 43.2 | 78.7 | 65.1 | |
| SAM 3 Detector + T-by-D as tracker2025.11 | 27.2 | 84 | 32.4 | 56.2 | 46.6 | 68.4 | 60.6 | |
| LLMDet + SAM 3 TrackerZero-shot=true2025.11 | 0.3 | 2 | 14.1 | 16.4 | 3.8 | 71.4 | 36.9 | |
| GLEEPrompting strategy=one NP at a time, Zero-shot=true2025.11 | 0.1 | 14 | 0.5 | 4.7 | 1.4 | 16 | 13.3 | |
| GLEEPrompting strategy=all NPs at once, Zero-shot=true2025.11 | 0 | 15 | 0.2 | 4.1 | 1.4 | 13.1 | 8.4 |