3D Semantic Segmentation on ScanNet200 (test)
24.6mIoU (f)Chorus
Evaluation results
| Method | Links | ||||||
|---|---|---|---|---|---|---|---|
| ChorusTraining Source=SN, SN++, MP3D, #Training Scenes=x2.922025.12 | 24.6 | — | — | — | — | 47.7 | |
| ChorusTraining Source=SN2025.12 | 22.4 | — | — | — | — | 45.8 | |
| SceneSplatTraining Source=SN, SN++, MP3D, #Training Scenes=x2.922025.12 | 21.4 | — | — | — | — | 38.7 | |
| SceneSplatTraining Source=SN2025.12 | 18.9 | — | — | — | — | 31.7 | |
| Mosaic3DTraining Source=SN, SN++, MP3D, ARKitS, S3D, #Training Scenes=x24.32025.12 | 15.7 | — | — | — | — | 28.3 | |
| Mosaic3DTraining Source=SN2025.12 | 13 | — | — | — | — | 24.5 | |
| MPECOpen-Vocabulary=true2025.04 | 10.8 | — | — | — | — | 27.4 | |
| RegionPLCTraining Source=SN2025.12 | 9.2 | — | — | — | — | 16.4 | |
| RegionPLCOpen-Vocabulary=true2025.04 | 9.1 | — | — | — | — | 17.3 | |
| OV3DOpen-Vocabulary=true2025.04 | 8.7 | — | — | — | — | — | |
| OV3DTraining Source=SN2025.12 | 8.7 | — | — | — | — | — | |
| OpenScene-3DOpen-Vocabulary=true, reproduced_by=OV3D2025.04 | 7.3 | — | — | — | — | — | |
| OpenSceneTraining Source=SN, #Training Scenes=x12025.12 | 6.4 | — | — | — | — | 12.2 | |
| PLAOpen-Vocabulary=true2025.04 | 1.8 | — | — | — | — | 3.1 | |
| PLATraining Source=SN2025.12 | 1.8 | — | — | — | — | 3.1 | |
| BFANetVenue=CVPR’252026.01 | — | 36 | 55.3 | 29.3 | 19.3 | — | |
| CeCoVenue=CVPR’232026.01 | — | 34 | 55.1 | 24.7 | 18.1 | — | |
| G2PVenue=-2026.01 | — | 35.5 | 55.3 | 28.6 | 18.7 | — | |
| MinkUNetVenue=CVPR’192026.01 | — | 25.3 | 46.3 | 15.4 | 10.6 | — | |
| MinkUNet (fine-tune)Backbone=MinkUNet, Training Data=ARKit LabelMaker SN200 -> ScanNet2002024.10 | — | 27.4 | 49 | 19.4 | 9.4 | — | |
| MinkUNet (vanilla)Backbone=MinkUNet, Training Data=ScanNet2002024.10 | — | 25.3 | 46.3 | 15.4 | 10.2 | — | |
| OctFormerVenue=TOG’232026.01 | — | 32.6 | 53.9 | 26.5 | 13.1 | — | |
| PonderV2Venue=CVPR’242026.01 | — | 34.6 | 55.2 | 27 | 17.5 | — | |
| PPTBackbone=PTv3, Training Data=ScanNet200 + S3DIS + Structure3D -> ScanNet2002024.10 | — | 39.3 | 59.2 | 33 | 21.6 | — | |
| PPT (joint-training)Backbone=PTv3, Training Data=ScanNet + ScanNet200 + ScanNet++ + Structure3D + ARKit LabelMaker2024.10 | — | 41.4 | 61 | 32.2 | 27.1 | — | |
| PTv3 (fine-tune)Backbone=PTv3, Training Data=ARKit LabelMaker -> ScanNet2002024.10 | — | 38.4 | 58.2 | 30.9 | 22.2 | — | |
| PTv3 (vanilla)Backbone=PTv3, Training Data=ScanNet2002024.10 | — | 37.8 | — | — | — | — |