Semantic Segmentation on nuScenes (val)
81.2mIoU (Segmentation)PPT + PTv3
Evaluation results
| Method | Links | ||||||||||||||||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| PPT + PTv3Params.=46.3M, Training Protocol=supervised pre-training (fine-tuned)2023.08 | 81.2 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| PTv3Params.=46.2M2023.08 | 80.4 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| PPT + SparseUNetParams.=41.0M, Training Protocol=supervised pre-training (fine-tuned)2023.08 | 78.6 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| PPT + SparseUNetParams.=41.0M, Training Protocol=supervised joint training2023.08 | 78.5 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| SphereFormerParams.=32.3M2023.08 | 78.4 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| Cylinder3DParams.=26.1M2023.08 | 76.1 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| SparseUNetParams.=39.2M2023.08 | 73.3 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| DITRModality=2D-3D fusion2025.03 | 0.842 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| LitePT-S#Params=12.7M2025.12 | 0.822 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | 88.1 | — | — | — | — | |
| UtoniaLearn.=157.7M, Pct.=100%, Evaluation Protocol=f.t.2026.03 | 0.822 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | 88.3 | 94.8 | — | — | — | |
| Volt-B#Params=94.3M, Joint training=true2026.04 | 0.822 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| ConcertoLearn.=157.7M, Pct.=100%, Evaluation Protocol=f.t.2026.03 | 0.82 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | 88.1 | 94.6 | — | — | — | |
| NOMAEProtocol=Fine-Tuning, Additional Data=false2025.12 | 0.818 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | 87.7 | — | — | — | — | |
| DOSProtocol=Fine-Tuning, Additional Data=true2025.12 | 0.818 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | 87.8 | — | — | — | — | |
| NOMAEBackbone=Point Transformer v3, Evaluation Protocol=fine-tuning2026.03 | 0.818 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| Volt-S#Params=23.7M, Joint training=true2026.04 | 0.818 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| SonataModality=3D-only2025.03 | 0.817 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| SonataProtocol=Fine-Tuning, Additional Data=true2025.12 | 0.817 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | 87.9 | — | — | — | — | |
| SonataEvaluation Protocol=full fine-tuning2025.03 | 0.817 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | 87.9 | 95 | — | — | — | |
| SonataLearn.=124.8M, Pct.=100%, Evaluation Protocol=f.t.2026.03 | 0.817 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | 87.9 | 95 | — | — | — | |
| DoReMi2025.11 | 0.817 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | 87.4 | 94.8 | — | — | — | |
| Flash3Dconstraint=Same memory costs as PTv32024.12 | 0.815 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| DOSProtocol=Fine-Tuning, Additional Data=false2025.12 | 0.815 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | 87.6 | — | — | — | — | |
| DOSBackbone=Point Transformer v3, Evaluation Protocol=fine-tuning2026.03 | 0.815 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| VaViT-Btest-time augmentation=false, reproduced results=false2026.05 | 0.813 | — | — | 0.836 | 79.6 | 54.4 | 95.2 | 93.9 | 65.4 | 88.8 | 72 | 74.1 | 87.3 | 97.1 | 75.3 | 77 | 76.2 | 91.3 | 89.7 | — | — | — | — | — | |
| PTv3 + PPTYear=20242023.12 | 0.812 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| PTv3Training Protocol=Multi-dataset joint training (PPT [92])2023.12 | 0.812 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| PTv3 + PPTYear=20242023.08 | 0.812 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| Flash3Dconstraint=Same parameters as PTv32024.12 | 0.812 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| PPTSupervision=supervised2025.03 | 0.812 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | 87.5 | 94.8 | — | — | — | |
| Sonatareproduction=reproduced from official repository2025.11 | 0.812 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | 87.7 | 94.8 | — | — | — | |
| PTv3/PPT#Params=46.3M, Joint training=true2026.04 | 0.812 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| RelFlexformer + PointRoPECategory=Performer, Base Model=PTv3, Position Encoding=PointRoPE2026.05 | 0.812 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| PointINSBackbone=Point Transformer v3, Evaluation Protocol=fine-tuning2026.03 | 0.811 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| Volt-S#Params=23.7M, Joint training=false2026.04 | 0.81 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| PTv3 + M3NetYear=20242023.12 | 0.809 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| 2DPASSCategory=MLP/CNN/SparseConv2026.05 | 0.808 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| D-DITRProtocol=Fine-Tuning, Additional Data=false2025.12 | 0.807 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| LitePTtest-time augmentation=false, reproduced results=true2026.05 | 0.807 | — | — | 0.828 | 79.2 | 52.3 | 95.7 | 94.2 | 64.4 | 84.8 | 68.9 | 72.9 | 87.7 | 97 | 76.7 | 77.3 | 76.8 | 91.1 | 89.6 | — | — | — | — | — | |
| Serialized Point Mamba2024.07 | 0.806 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| LCPSModality=2D-3D fusion2025.03 | 0.805 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| PTv3Year=20242023.12 | 0.804 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| PTv3Training Protocol=Trained from scratch2023.12 | 0.804 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| PTv3Year=20242023.08 | 0.804 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| PTv32024.12 | 0.804 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| PTv3Modality=3D-only2025.03 | 0.804 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| PTv3#Params=46.1M2025.12 | 0.804 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | 87.2 | — | — | — | — | |
| PTv3Protocol=Supervised, Additional Data=false2025.12 | 0.804 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | 87.2 | — | — | — | — | |
| PTv3Supervision=supervised2025.03 | 0.804 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | 87.2 | 94.7 | — | — | — | |
| PTv3Learn.=124.8M, Pct.=100%2026.03 | 0.804 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | 87.2 | 94.7 | — | — | — | |
| PTv32025.11 | 0.804 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | 87.2 | 94.7 | — | — | — | |
| PTv3#Params=46.1M, Joint training=false2026.04 | 0.804 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| PTv3Category=Transformer2026.05 | 0.804 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| PambaCategory=Mamba2026.05 | 0.804 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| PTv3Params=46.1M, Resolution=τ*2026.05 | 0.804 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | 86.2 | — | — | — | — | |
| PTv3Evaluation protocol=supervised2026.03 | 0.803 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| PTv3 (sup.)Backbone=Point Transformer v3, Evaluation Protocol=Supervised2026.03 | 0.803 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| RelFlexformerCategory=Performer, Base Model=PTv32026.05 | 0.803 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| PTv2Year=20222023.12 | 0.802 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| PTv2Training Protocol=Trained from scratch2023.12 | 0.802 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| PTv22024.07 | 0.802 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| PTv2Year=20222023.08 | 0.802 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| PTv22024.12 | 0.802 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| PTv2Modality=3D-only2025.03 | 0.802 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| PTv2#Params=12.8M2025.12 | 0.802 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| PTv2#Params=12.8M, Joint training=false2026.04 | 0.802 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| PTv2Category=Transformer2026.05 | 0.802 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| NOMAEBackbone=Point Transformer v3, Evaluation Protocol=decoder tuning2026.03 | 0.801 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| LSK3DNet#Params=28.8M, Joint training=false2026.04 | 0.801 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| MSeg3DModality=2D-3D fusion2025.03 | 0.8 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| PointINSBackbone=Point Transformer v3, Evaluation Protocol=decoder tuning2026.03 | 0.8 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| ConDaFormerTesting-time augmentations=true2023.12 | 0.799 | — | — | 0.829 | 79.2 | 47.8 | 95.3 | 94.1 | 57.2 | 86.6 | 69 | 73.6 | 87.4 | 97.1 | 74.8 | 76.6 | 76 | 91.2 | 89.6 | — | — | — | — | — | |
| PTv3Modality=3D-only, Evaluation setup=reproduced2025.03 | 0.799 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| SonataBackbone=Point Transformer v3, Evaluation Protocol=fine-tuning2026.03 | 0.796 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| SphereFormerrotation and translation testing-time augmentations=true2023.03 | 0.795 | — | — | 0.811 | 78.7 | 46.7 | 95.2 | 93.7 | 54 | 88.9 | 68 | 74.2 | 86.2 | 97.2 | 74.3 | 76.3 | 75.8 | 91.4 | 89.7 | — | — | — | — | — | |
| SphereFormerTesting-time augmentations=true2023.12 | 0.795 | — | — | 0.811 | 78.7 | 46.7 | 95.2 | 93.7 | 54 | 88.9 | 68 | 74.2 | 86.2 | 97.2 | 74.3 | 76.3 | 75.8 | 91.4 | 89.7 | — | — | — | — | — | |
| SphereFormerType=Voxel2023.03 | 0.795 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| 2DPASSTesting-time augmentations=true2023.12 | 0.794 | — | — | 0.822 | 78.8 | 49.6 | 95.6 | 93.6 | 60 | 84.1 | 67.5 | 72.6 | 88.1 | 96.8 | 72.8 | 76.2 | 76.5 | 89.4 | 87.2 | — | — | — | — | — | |
| 2DPASSModality=2D-3D fusion2025.03 | 0.794 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| DOSBackbone=Point Transformer v3, Evaluation Protocol=decoder tuning2026.03 | 0.792 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| MinkUNet + M3NetYear=20242023.12 | 0.79 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| MinkUNet + M3NetCategory=MLP/CNN/SparseConv, Enhancement=M3Net2026.05 | 0.79 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| OA-CNNsYear=20242023.12 | 0.789 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| OA-CNNsYear=20242023.08 | 0.789 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| 4D-FormerModality=2D-3D fusion2025.03 | 0.789 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| OA-CNNsCategory=MLP/CNN/SparseConv2026.05 | 0.789 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| PTv3test-time augmentation=false, reproduced results=true2026.05 | 0.789 | — | — | 0.822 | 79.7 | 49.7 | 95.6 | 93.6 | 46 | 84.6 | 70.1 | 72.9 | 84 | 96.9 | 74.7 | 76.3 | 75.5 | 90.8 | 89.1 | — | — | — | — | — | |
| WaffleIronSupervision level=full supervision2025.07 | 0.787 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| PSABackbone=Point Transformer v3, Evaluation Protocol=fine-tuning2026.03 | 0.787 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| WI-768test-time augmentation=false, reproduced results=false2026.05 | 0.787 | — | — | 0.834 | 79.2 | 53.2 | 92.5 | 88.1 | 50.4 | 87.8 | 70.3 | 73.5 | 84.1 | 96.9 | 73.3 | 75.5 | 75.4 | 89.9 | 86.5 | — | — | — | — | — | |
| MinkUNet + PPTYear=20242023.12 | 0.786 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| MinkUNet + PPTYear=20242023.08 | 0.786 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| MinkUNet + PPTCategory=MLP/CNN/SparseConv, Enhancement=PPT2026.05 | 0.786 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| FlatFormer-Stest-time augmentation=false, reproduced results=true2026.05 | 0.786 | — | — | 0.799 | 78.8 | 45.4 | 94.5 | 93 | 58.1 | 81.5 | 67.2 | 71.5 | 85.2 | 96.8 | 74.5 | 75.9 | 75.4 | 90.8 | 89.2 | — | — | — | — | — | |
| SphereFormer2023.03 | 0.784 | — | — | 0.812 | 77.7 | 43.8 | 94.5 | 93.1 | 52.4 | 86.9 | 65.4 | 73.4 | 85.3 | 97 | 73.4 | 75.4 | 75 | 91 | 89.2 | — | — | — | — | — | |
| SphereFormerYear=20232023.12 | 0.784 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| SphereFormerTraining Protocol=Trained from scratch2023.12 | 0.784 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| SphereFormer2024.07 | 0.784 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| SphereFormerYear=20232023.08 | 0.784 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| SphereFormerApproach Type=3D Voxel Based2023.11 | 0.784 | — | — | 0.812 | 77.7 | 43.8 | 94.5 | 93.1 | 52.4 | 86.9 | 65.4 | 73.4 | 85.3 | 97 | 73.4 | 75.4 | 75 | 91 | 89.2 | — | — | — | — | — |