Open-vocabulary video instance segmentation on YT-Temporal-1B SA-Co VEval (test)
71.2cgF1Human
Evaluation results
| Method | Links | ||
|---|---|---|---|
| Human2025.11 | 71.2 | 78.4 | |
| SAM 32025.11 | 50.8 | 69.9 | |
| SAM 3 Detector + T-by-D2025.11 | 47.6 | 68.2 | |
| LLMDet + SAM 3 TrackerZero-shot=true2025.11 | 8 | 37.9 | |
| GLEEProcessing mode=one NP at a time, Zero-shot=true2025.11 | 2.2 | 18.9 | |
| GLEEProcessing mode=all NPs at once, Zero-shot=true2025.11 | 1.6 | 16.7 |