Open-Vocabulary Semantic Segmentation on ScanNet++ OpenLex3D
0.4S ScoreHOV-SG
Evaluation results
| Method | Links | ||||||
|---|---|---|---|---|---|---|---|
| HOV-SGInput=ground-truth poses and depth2026.05 | 0.4 | 0.02 | 0.04 | 0.16 | 0.08 | 0.3 | |
| LEXI-SGInput=ground-truth poses and depth2026.05 | 0.4 | 0.01 | 0.04 | 0.19 | 0.12 | 0.24 | |
| OpenMask3DInput=ground-truth poses and depth2026.05 | 0.27 | 0.01 | 0.03 | 0.29 | 0.13 | 0.27 | |
| ConceptGraphsInput=ground-truth poses and depth2026.05 | 0.26 | 0.02 | 0.05 | 0.1 | 0.13 | 0.44 |