Cross-modal Retrieval on ScanNet N=3072 (test)
46.1R@1Omnivorous ViT-B/14
Evaluation Results
| Method | Links | ||||
|---|---|---|---|---|---|
| Omnivorous ViT-B/14feature type=gap2026.02 | 46.1 | 71.4 | 57.7 | 2 | |
| Omnivorous ViT-B/14feature type=tok2026.02 | 30.2 | 55.8 | 42.2 | 5.3 | |
| DINOv2 ViT-B/14feature type=gap2026.02 | 4.6 | 10.8 | 8.1 | 401.8 | |
| DINOv2 ViT-B/14feature type=tok2026.02 | 3.9 | 9 | 6.9 | 382.5 |