Open-vocabulary Segmentation on ScanNet Target View
51.3LSeg mIoUC3G
Evaluation results
| Method | Links | ||||||
|---|---|---|---|---|---|---|---|
| C3GFeed Forward=true, Input Pose=false2025.12 | 51.3 | 78.3 | 36.9 | 67.5 | — | — | |
| LSeg / MaskCLIP*Feed Forward=false2025.12 | 50.6 | 79.7 | 34.1 | 66.7 | — | — | |
| LSMFeed Forward=true, Input Pose=false2025.12 | 50.3 | 79.3 | 28.6 | 50.5 | — | — | |
| Feature-3DGSFeed Forward=false, Input Pose=true2025.12 | 37.9 | 64.4 | 35.3 | 66.3 | — | — | |
| CF3Feed Forward=false, Input Pose=true2025.12 | 37.6 | 65.7 | 33.6 | 63.4 | — | — | |
| Feature-3DGSRecon. Time (SfM)=20.52s, Recon. Time (Per-Scene)=18mins2026.04 | — | — | — | — | 42.23 | 71.74 | |
| LSegRecon. Time (SfM)=N/A, Recon. Time (Per-Scene)=N/A2026.04 | — | — | — | — | 48.19 | 79.27 | |
| LSMRecon. Time (Per-Scene)=0.108s2026.04 | — | — | — | — | 50.78 | 76.86 | |
| NeRF-DFFRecon. Time (SfM)=20.52s, Recon. Time (Per-Scene)=1min2026.04 | — | — | — | — | 40.37 | 67.55 | |
| UniSplatRecon. Time (Per-Scene)=0.041s2026.04 | — | — | — | — | 56.25 | 83.34 |