Open-vocabulary video instance segmentation on SA-V SA-Co/VEval (test)
58.5cgF1Human
Evaluation results
| Method | Links | |||||||
|---|---|---|---|---|---|---|---|---|
| Human2025.11 | 58.5 | 72.3 | — | — | — | — | — | |
| Human2025.11 | 53.1 | 70.5 | — | — | — | — | — | |
| Human2025.11 | 53.1 | 70.5 | 0.87 | 60.8 | 55.5 | 90.2 | 69.1 | |
| SAM 32025.11 | 36.4 | 63.6 | — | — | — | — | — | |
| SAM 32025.11 | 30.3 | 58 | — | — | — | — | — | |
| SAM 32025.11 | 30.3 | 58 | 0.69 | 43.7 | 40.9 | 83.4 | 56.8 | |
| SAM 3 Detector + T-by-D2025.11 | 29.7 | 60 | — | — | — | — | — | |
| SAM 3temporal disambiguation=false2025.11 | 27.1 | 55.9 | 0.63 | 42.9 | 37.9 | 83.4 | 57.1 | |
| SAM 3 Detector + T-by-D2025.11 | 25.7 | 55.7 | — | — | — | — | — | |
| SAM 3 Detector + T-by-D as tracker2025.11 | 25.7 | 55.7 | 0.72 | 35.8 | 40.2 | 78.3 | 53.2 | |
| LLMDet + SAM 3 TrackerZero-shot=true2025.11 | 2.3 | 30.1 | — | — | — | — | — | |
| LLMDet + SAM 3 TrackerZero-shot=true2025.11 | 2.3 | 30.1 | 0.15 | 14.7 | 12 | 76 | 28.5 | |
| LLMDet + SAM 3 TrackerZero-shot=true2025.11 | 0.3 | 18.6 | — | — | — | — | — | |
| GLEEProcessing mode=all NPs at once, Zero-shot=true2025.11 | 0.1 | 8.7 | — | — | — | — | — | |
| GLEEProcessing mode=one NP at a time, Zero-shot=true2025.11 | 0.1 | 11.8 | — | — | — | — | — | |
| GLEEProcessing mode=one NP at a time, Zero-shot=true2025.11 | 0.1 | 5.6 | — | — | — | — | — | |
| GLEEPrompting strategy=all NPs at once, Zero-shot=true2025.11 | 0.1 | 8.7 | 0.22 | 0.6 | 2.8 | 30.9 | 7.1 | |
| GLEEPrompting strategy=one NP at a time, Zero-shot=true2025.11 | 0.1 | 11.8 | 0.05 | 2.2 | 3.4 | 43.1 | 14.9 | |
| GLEEProcessing mode=all NPs at once, Zero-shot=true2025.11 | 0 | 4.7 | — | — | — | — | — |