Open-Vocabulary 3D Segmentation on ScanNet (40 unseen scenes)
56.77mIoUUniSem
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| UniSemType=Feed-Forward, Recon. Time=0.162s2026.03 | 56.77 | 83.45 | |
| SpatialSplatType=Feed-Forward, Recon. Time=0.071s2026.03 | 55.87 | 79.24 | |
| Uni3RType=Feed-Forward, Recon. Time=0.162s2026.03 | 54.74 | 81.98 | |
| LSMType=Feed-Forward, Recon. Time=0.108s2026.03 | 50.78 | 76.86 | |
| LSegType=Per-Scene2026.03 | 48.19 | 79.27 | |
| Feat3DGSType=Per-Scene, Recon. Time=18m21s2026.03 | 42.23 | 71.74 | |
| DFFType=Per-Scene, Recon. Time=1m21s2026.03 | 40.37 | 67.55 |