Zero-shot Object Detection on LVIS rare V1 (undetected failure samples)
11.55Instance MetricGreedy
Evaluation Results
| Method | Links | ||||||
|---|---|---|---|---|---|---|---|
| GreedyBackbone=Grounding DINO, Steps=1002026.06 | 11.55 | 18.86 | 17.84 | 30.15 | 10,100 | 0.18 | |
| PhaseWinBackbone=Grounding DINO, Steps=1002026.06 | 9.42 | 14.38 | 15.52 | 24.72 | 1,509.1 | 1.03 | |
| PhaseWinBackbone=Grounding DINO, Steps=502026.06 | 7.87 | 12.86 | 13.09 | 17.04 | 348.4 | 3.76 | |
| RISEBackbone=Grounding DINO2026.06 | 7.03 | 11.84 | 13.12 | 18.73 | 5,000 | 0.26 | |
| HSICBackbone=Grounding DINO2026.06 | 5.16 | 9.2 | 11.68 | 13.48 | 1,536 | 0.76 | |
| SSGrad-CAM++Backbone=Grounding DINO2026.06 | 2.92 | 6.8 | 8.97 | 5.24 | — | — | |
| Grad-CAMBackbone=Grounding DINO2026.06 | 2.91 | 6.89 | 9.01 | 5.43 | — | — | |
| ODAMBackbone=Grounding DINO2026.06 | 2.83 | 7.16 | 8.51 | 4.68 | — | — |