3D Instance Segmentation on S3DIS (Area 5)
82.8mAP@50% IoUEvObj (Diffusion)
Evaluation Results
| Method | Links | |||||||
|---|---|---|---|---|---|---|---|---|
| EvObj (Diffusion)Supervision level=Unsupervised2026.05 | 82.8 | — | — | 60.6 | — | — | 92.1 | |
| EvObj (VAE)Supervision level=Unsupervised2026.05 | 77.1 | — | — | 55.1 | — | — | 86.2 | |
| Mask3D2024.11 | 71.9 | 63.7 | 74.3 | — | — | — | — | |
| BFL2025.02 | 71.9 | — | — | — | — | — | 77.8 | |
| MSTA3D2024.11 | 70 | 70.1 | 80.6 | — | — | — | — | |
| QueryFormer2024.11 | 69.9 | 72.2 | 70.5 | — | — | — | — | |
| QueryFormer2025.02 | 69.9 | — | — | — | — | — | — | |
| Ours2023.09 | 69.1 | — | — | — | — | — | 75.7 | |
| MAFTInput=Mesh Sampled Point Cloud, Uses additional data=false2024.01 | 69.1 | — | — | — | — | — | 75.7 | |
| MAFT2024.11 | 69.1 | — | — | — | — | — | — | |
| Maft2025.02 | 69.1 | — | — | — | — | — | 75.7 | |
| MAFTVenue=ICCV 232026.02 | 69.1 | — | — | — | — | — | — | |
| SoftGroupevaluation_protocol=Area 52022.03 | 68.9 | 69.8 | 75.3 | 54.4 | 69.3 | 71.7 | — | |
| OneFormer3DVenue=CVPR 24, Pre-trained on ScanNet V2=true2026.02 | 68.5 | — | — | — | — | — | — | |
| Mask3D2023.09 | 68.4 | — | — | — | — | — | 75.2 | |
| Mask3D2025.02 | 68.4 | — | — | — | — | — | 75.2 | |
| SSTNetevaluation_protocol=Area 52022.03 | 67.8 | 73.4 | 73.5 | 54.1 | — | — | — | |
| SoftGroup++subsampling=none2022.09 | 67.8 | 67.6 | 73.8 | 50.9 | — | — | — | |
| ISBNet2023.03 | 67.5 | 72 | 70.5 | 56.3 | 70 | 70.7 | — | |
| SPFormerevaluation_split=Area 52022.11 | 66.8 | 67.1 | 72.8 | — | — | — | — | |
| SPFormer2023.09 | 66.8 | — | — | — | — | — | — | |
| SPFormer2024.11 | 66.8 | 67.1 | 72.8 | — | — | — | — | |
| SPFormer2025.02 | 66.8 | — | — | — | — | — | — | |
| SPFormerVenue=AAAI 232026.02 | 66.8 | — | — | — | — | — | — | |
| PBNet2024.11 | 66.4 | 65.4 | 74.9 | — | — | — | — | |
| GrabS-VAESupervision level=Unsupervised2026.05 | 66.2 | — | — | 46.4 | — | — | 73.8 | |
| SoftGroupevaluation_split=Area 52022.11 | 66.1 | 66.6 | 73.6 | — | — | — | — | |
| SoftGroup2023.03 | 66.1 | 66.6 | 73.6 | 51.6 | 66.1 | 68 | — | |
| SoftGroup2023.09 | 66.1 | — | — | — | — | — | — | |
| SoftGroupsubsampling=1/42022.09 | 66.1 | 66.6 | 73.6 | 51.6 | — | — | — | |
| SoftGroup2024.11 | 66.1 | 66.6 | 73.6 | — | — | — | — | |
| SoftGroup2025.02 | 66.1 | — | — | — | — | — | — | |
| TD3D2024.11 | 65.1 | 64.8 | 74.4 | — | — | — | — | |
| MaskGroup2022.03 | 65 | 64.7 | 62.9 | — | — | — | — | |
| MaskGroup2023.09 | 65 | — | — | — | — | — | — | |
| MaskGroup2025.02 | 65 | — | — | — | — | — | — | |
| MaskGroupEvaluation Split=Area 52022.07 | 65 | 64.7 | 62.9 | — | — | — | — | |
| PointGroupevaluation_protocol=Area 52022.03 | 64 | 69.2 | 69.6 | — | — | — | — | |
| UnScene3DSupervision level=Unsupervised2026.05 | 63.4 | — | — | 42.6 | — | — | 80.3 | |
| 3D-MPA2020.03 | 63.1 | 58 | — | — | — | — | — | |
| Spherical Mask2023.12 | 60.5 | 76.3 | 71.3 | 72.3 | — | — | — | |
| SSTNetevaluation_split=Area 52022.11 | 59.3 | 64.2 | 65.5 | — | — | — | — | |
| SSTNet2023.03 | 59.3 | 64.2 | 65.6 | 42.7 | — | — | — | |
| SSTNet2023.09 | 59.3 | — | — | — | — | — | — | |
| SSTNet2022.09 | 59.3 | 64.2 | 65.5 | 42.7 | — | — | — | |
| SSTNet2025.02 | 59.3 | — | — | — | — | — | — | |
| GrabS-DiffusionSupervision level=Unsupervised2026.05 | 58 | — | — | 44.2 | — | — | 62.6 | |
| Mask3DInput=Mesh Sampled Point Cloud, Uses additional data=true2024.01 | 57.8 | — | — | — | — | — | 77.2 | |
| PointGroupEvaluation Split=Area 52020.04 | 57.8 | 62.1 | 61.9 | — | — | — | — | |
| PointGroup2022.03 | 57.8 | 62.1 | 61.9 | — | — | — | — | |
| PointGroupevaluation_split=Area 52022.11 | 57.8 | 42.4 | 55.3 | — | — | — | — | |
| PointGroup2023.03 | 57.8 | 62.1 | 61.9 | — | — | — | — | |
| PointGroup2023.09 | 57.8 | — | — | — | — | — | — | |
| PointGroup2022.09 | 57.8 | 62.1 | 61.9 | — | — | — | — | |
| PointGroup2025.02 | 57.8 | — | — | — | — | — | — | |
| PointGroupVenue=CVPR 202026.02 | 57.8 | — | — | — | — | — | — | |
| PointGroup2021.08 | 57.8 | 62.1 | 61.9 | — | — | — | — | |
| PointGroupEvaluation Split=Area 52022.07 | 57.8 | 62.1 | 61.9 | — | — | — | — | |
| QueryFormerInput=Mesh Sampled Point Cloud, Uses additional data=false2024.01 | 57.7 | — | — | — | — | — | — | |
| QueryFormer2023.12 | 57.7 | 72.2 | 70.5 | 69.9 | — | — | — | |
| 3D-BoNet2020.03 | 57.5 | 40.2 | — | — | — | — | — | |
| Sonata (full)Params=124.8M, Learn.=124.8M, Pct.=100%, Evaluation Protocol=full fine-tuning2025.03 | 57.4 | — | — | 45.5 | — | — | 63.8 | |
| Sonata (dec.)Params=124.8M, Learn.=16.3M, Pct.=13%, Evaluation Protocol=decoder probing2025.03 | 57.1 | — | — | 45.1 | — | — | 63.7 | |
| Mask3DInput=Mesh Sampled Point Cloud, Uses additional data=false2024.01 | 56.6 | — | — | — | — | — | 75.2 | |
| LaSSM2026.02 | 56.5 | — | — | — | — | — | — | |
| Mask3D2023.12 | 56.5 | 70.7 | 68.7 | 69.3 | — | — | — | |
| ISBNet2023.12 | 56.3 | 72 | 70.5 | 67.5 | — | — | — | |
| SGIFormerVenue=TCSVT 252026.02 | 56.2 | — | — | — | — | — | — | |
| ASIS2020.03 | 55.3 | 42.4 | — | — | — | — | — | |
| PBNetVenue=ICCV 232026.02 | 53.5 | — | — | — | — | — | — | |
| PBNetEvaluation Split=Area 52022.07 | 53.5 | 65.4 | 74.9 | — | — | — | — | |
| PBNet2023.12 | 53.5 | 65.4 | 74.9 | 66.4 | — | — | — | |
| SoftGroupInput=Mesh Sampled Point Cloud, Uses additional data=true2024.01 | 51.6 | — | — | — | — | — | — | |
| SoftGroupVenue=CVPR 222026.02 | 51.6 | — | — | — | — | — | — | |
| SoftGroupEvaluation Split=Area 52022.07 | 51.6 | 66.6 | 73.6 | — | — | — | — | |
| SoftGroup2023.12 | 51.6 | 66.6 | 73.6 | 66.1 | — | — | — | |
| PPT (sup.)Params=124.8M, Learn.=124.8M, Pct.=100%, Evaluation Protocol=full fine-tuning2025.03 | 51.2 | — | — | 39.7 | — | — | 57.5 | |
| MSC (full)Params=124.8M, Learn.=124.8M, Pct.=100%, Evaluation Protocol=full fine-tuning2025.03 | 50.5 | — | — | 38.1 | — | — | 56.3 | |
| Part2ObjectSupervision level=Unsupervised2026.05 | 50.5 | — | — | 30 | — | — | 76.4 | |
| PointGroupParams=124.8M, Learn.=124.8M, Pct.=100%2025.03 | 49.4 | — | — | 37.8 | — | — | 55.7 | |
| TD3DVenue=WACV 242026.02 | 48.6 | — | — | — | — | — | — | |
| LAM3C*VGPC Pretraining Data=49k, Evaluation Protocol=Full-FT, Backbone=PTv3 (Large), Pre-training steps=434k2025.12 | 47.2 | — | — | — | — | — | — | |
| LAM3C*Real Pretraining Data=15k, VGPC Pretraining Data=49k, Evaluation Protocol=Full-FT, Backbone=PTv3 (Large), Pre-training steps=434k2025.12 | 46.8 | — | — | — | — | — | — | |
| LAM3CVGPC Pretraining Data=49k, Evaluation Protocol=Full-FT, Backbone=PTv3 (Base)2025.12 | 45.7 | — | — | — | — | — | — | |
| SonataReal Pretraining Data=18k, Synth Pretraining Data=121k, Evaluation Protocol=Full-FT, Backbone=PTv3 (Base)2025.12 | 45.5 | — | — | — | — | — | — | |
| ODIN-ResNet50Input=RGBD Point Cloud, Uses additional data=true2024.01 | 44.7 | — | — | — | — | — | 67.5 | |
| LAM3CVGPC Pretraining Data=16k, Evaluation Protocol=Full-FT, Backbone=PTv3 (Base)2025.12 | 43.3 | — | — | — | — | — | — | |
| ODIN-Swin-BInput=RGBD Point Cloud, Uses additional data=true2024.01 | 43 | — | — | — | — | — | 70 | |
| SSTNet2021.08 | 42.7 | 64.2 | 65.5 | — | — | — | — | |
| SSTNetEvaluation Split=Area 52022.07 | 42.7 | 64.2 | 65.5 | — | — | — | — | |
| SSTNet2023.12 | 42.7 | 64.2 | 65.6 | 59.3 | — | — | — | |
| Sonata (all real)Real Pretraining Data=15k, Evaluation Protocol=Full-FT, Backbone=PTv3 (Base)2025.12 | 41.5 | — | — | — | — | — | — | |
| Mask3DInput=RGBD Point Cloud, Uses additional data=true2024.01 | 41.3 | — | — | — | — | — | 66.1 | |
| Sonata (ScanNet)Real Pretraining Data=1k, Evaluation Protocol=Full-FT, Backbone=PTv3 (Base)2025.12 | 40.8 | — | — | — | — | — | — | |
| Mask3DInput=RGBD Point Cloud, Uses additional data=false2024.01 | 40.7 | — | — | — | — | — | 64.2 | |
| UnScene3D2023.03 | 40.3 | — | — | 21.4 | — | — | 52.6 | |
| PPTReal Pretraining Data=1k, Synth Pretraining Data=21k, Evaluation Protocol=Full-FT, Backbone=PTv3 (Base)2025.12 | 39.7 | — | — | — | — | — | — | |
| PTv3Evaluation Protocol=Full-FT, Backbone=PTv3 (Base)2025.12 | 39.6 | — | — | — | — | — | — | |
| MSCReal Pretraining Data=7k, Evaluation Protocol=Full-FT, Backbone=PTv3 (Base)2025.12 | 38.1 | — | — | — | — | — | — | |
| Sonata (lin.)Params=124.8M, Learn.=<0.2M, Pct.=<0.2%, Evaluation Protocol=linear probing2025.03 | 36.6 | — | — | 26.1 | — | — | 45.8 |