3D Semantic Segmentation on ScanNet 200 70 classes
34.75mIoUEmbodiedSplat (FastSAM + OpenSeg)
Evaluation results
| Method | Links | ||
|---|---|---|---|
| EmbodiedSplat (FastSAM + OpenSeg)Inputs=RGB-D, Online / Offline=Online2026.03 | 34.75 | 52.36 | |
| OpenSceneInputs=Point Cloud, RGB-D, Online / Offline=Offline2026.03 | 33.84 | 50.06 | |
| EmbodiedSAMInputs=RGB-D, Online / Offline=Online2026.03 | 33.11 | 48.14 | |
| EmbodiedSplat (FastSAM + OpenSeg)Inputs=RGB, Online / Offline=Online2026.03 | 31.16 | 48.38 | |
| EmbodiedSplat (FastSAM + LSeg)Inputs=RGB, Online / Offline=Online2026.03 | 23.13 | 35.18 | |
| EmbodiedSplat (SAM + CLIP)Inputs=RGB, Online / Offline=Online2026.03 | 21.98 | 41.11 | |
| LUDVIGInputs=RGB, Online / Offline=Offline2026.03 | 21.23 | 39.87 | |
| Occam’s LGSInputs=RGB, Online / Offline=Offline2026.03 | 20.32 | 40.49 | |
| Dr. SplatInputs=RGB, Online / Offline=Offline2026.03 | 19.29 | 33.84 | |
| CF3Inputs=RGB, Online / Offline=Offline2026.03 | 18.23 | 30.11 |