Cross-Modal Scene Retrieval (Image to Text) on ScanNet
57.39Scene Category Recall@1CrossOver
Evaluation Results
| Method | Links | |||||||||
|---|---|---|---|---|---|---|---|---|---|---|
| CrossOverDirection=I → R2025.02 | 57.39 | 82.82 | 87.63 | 3 | 25 | 51 | 29.04 | 57.85 | 70.75 | |
| PointBindDirection=I → R2025.02 | 49.48 | 70.45 | 80.07 | 2 | 6 | 12 | 19.19 | 41.54 | 55.85 | |
| Inst. Baseline (Ours)Direction=I → R2025.02 | 49.14 | 71.48 | 80.07 | 8 | 28 | 46 | 28 | 62.33 | 72.62 | |
| ULIP-2Direction=I → R2025.02 | 41.92 | 57.73 | 61.86 | 1 | 2 | 8 | 19.48 | 42.18 | 56.69 |