Cross-Modal Scene Retrieval (Image to Point Cloud) on ScanNet
64.74Scene Category Recall@1CrossOver
Evaluation Results
| Method | Links | |||||||||
|---|---|---|---|---|---|---|---|---|---|---|
| CrossOverDirection=I → P2025.02 | 64.74 | 89.42 | 94.23 | 13 | 41 | 84 | 38.98 | 73.28 | 85 | |
| Inst. Baseline (Ours)Direction=I → P2025.02 | 42.95 | 70.19 | 81.09 | 13 | 35 | 60 | 46.37 | 79.68 | 88.43 | |
| PointBindDirection=I → P2025.02 | 13.78 | 24.36 | 42.95 | 2 | 5 | 7 | 20.03 | 40.68 | 57.01 | |
| ULIP-2Direction=I → P2025.02 | 7.37 | 25.96 | 43.27 | 0.04 | 1 | 3 | 16.77 | 41.53 | 55.54 |