3D Semantic Segmentation on ScanNet 10 classes
57.48mIoUEmbodiedSplat (FastSAM + LSeg)
Evaluation results
| Method | Links | ||
|---|---|---|---|
| EmbodiedSplat (FastSAM + LSeg)Inputs=RGB, Online / Offline=Online2026.03 | 57.48 | 77.63 | |
| EmbodiedSplat (FastSAM + OpenSeg)Inputs=RGB-D, Online / Offline=Online2026.03 | 57.41 | 82.45 | |
| OpenSceneInputs=Point Cloud, RGB-D, Online / Offline=Offline2026.03 | 54.56 | 80.45 | |
| EmbodiedSAMInputs=RGB-D, Online / Offline=Online2026.03 | 52.13 | 78.13 | |
| EmbodiedSplat (FastSAM + OpenSeg)Inputs=RGB, Online / Offline=Online2026.03 | 49.81 | 76.13 | |
| ReLaGSTraining-Free=true2026.03 | 47.17 | 66.08 | |
| THGSTraining-Free=true2026.03 | 46.38 | 64.74 | |
| VALATraining-Free=true2026.03 | 46.21 | 65.61 | |
| EmbodiedSplat (SAM + CLIP)Inputs=RGB, Online / Offline=Online2026.03 | 45.56 | 75.13 | |
| Occam’sTraining-Free=true2026.03 | 45.16 | 64.39 | |
| LAGATraining-Free=false2026.03 | 42.6 | 63.2 | |
| Occam’s LGSInputs=RGB, Online / Offline=Offline2026.03 | 42.14 | 70.28 | |
| LUDVIGInputs=RGB, Online / Offline=Offline2026.03 | 41.11 | 68.34 | |
| Dr. SplatInputs=RGB, Online / Offline=Offline2026.03 | 39.21 | 66.66 | |
| OpenGau.Training-Free=false2026.03 | 38.29 | 55.19 | |
| CF3Inputs=RGB, Online / Offline=Offline2026.03 | 38.14 | 65.13 | |
| LangSplatTraining-Free=false2026.03 | 8.4 | 22.06 |