Part Segmentation on ShapeNetPart
89mIoU (Instance)MCFT
Evaluation results
| Method | Links | ||||||||||||||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| MCFTPre-trained Encoder=Point-MAE2026.03 | 89 | 86.4 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| PointStack2022.05 | 87.2 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| MCFTPre-trained Encoder=Point-BERT2026.03 | 87.1 | 85.2 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| PointNeXt2023.12 | 87 | 85.2 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| CurveNet2022.05 | 86.8 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| CurveNet2023.12 | 86.8 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| ConDaFormer2023.12 | 86.8 | 84.6 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| MCFTPre-trained Encoder=ReCon2026.03 | 86.8 | 84.4 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| PointGPT-LBackbone=PointGPT-L, Trainable Parameters=339.39M2024.10 | 86.6 | 84.8 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| Point Transformer2020.12 | 86.6 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | 83.7 | — | |
| Point TransformerParams (M)=7.82022.11 | 86.6 | 83.7 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| StratifiedFormer2022.11 | 86.6 | 85.1 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| PTv12023.12 | 86.6 | 83.7 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| CurveNetInputs=2k, Param (M)=5.5, GFLOPs (G)=2.5, Train Time=56.9 h2026.01 | 86.6 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| GDANet2020.12 | 86.5 | 85 | 84.2 | 88 | 90.6 | 80.2 | 90.7 | 82 | 91.9 | 88.5 | 82.7 | 96.1 | 75.8 | 95.7 | 83.9 | 62.9 | 83.1 | 84.4 | — | — | — | — | |
| P2PBackbone=ConvNeXt-Large, Head=UPerNet2022.08 | 86.5 | 84.1 | 84.3 | 85.1 | 88.3 | 80.4 | 91.6 | 80.8 | 92.1 | 87.9 | 85.6 | 95.9 | — | 94.2 | 82.4 | 62.7 | 74.7 | 83.7 | 76.1 | — | — | — | |
| RECONTraining protocol=Self-Supervised (Full fine-tuning), #TP (M)=27.062024.08 | 86.4 | 84.66 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| DensePointinput=2k2019.09 | 86.4 | 84.2 | 84 | 85.4 | 90 | 79.2 | 91.1 | 81.6 | 91.5 | 87.5 | 84.7 | 95.9 | — | 94.6 | 82.9 | 64.6 | 76.8 | 83.7 | 74.3 | — | — | — | |
| DensePoint2020.12 | 86.4 | 84.2 | 84 | 85.4 | 90 | 79.2 | 91.1 | 81.6 | 91.5 | 87.5 | 84.7 | 95.9 | 74.3 | 94.6 | 82.9 | 64.6 | 76.8 | 83.7 | — | — | — | — | |
| KPConv2020.12 | 86.4 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | 85.1 | — | |
| KPConv2022.05 | 86.4 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| DRNet2022.05 | 86.4 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| KPConv2022.08 | 86.4 | 85.1 | 84.6 | 86.3 | 87.2 | 81.1 | 91.1 | 77.8 | 92.6 | 88.4 | 82.7 | 96.2 | — | 95.8 | 85.4 | 69 | 82 | 83.6 | 78.1 | — | — | — | |
| KPConv2022.11 | 86.4 | 85.1 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| KPConv2023.12 | 86.4 | 85.1 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| PCT2023.12 | 86.4 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| AdaptConv2023.12 | 86.4 | 83.4 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| InterpCNN2020.12 | 86.3 | 84 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| InterpCNN2020.12 | 86.3 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | 84 | — | |
| PRA-Net2022.05 | 86.3 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| HybridNet+MAPTraining Strategy=with Self-supervised pretraining, Parameters (M)=25.1, FLOPs (G)=12.92024.10 | 86.3 | 84.7 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| GaussFusion2026.07 | 86.3 | 84.43 | 85.2 | 84.2 | 87.7 | 80.4 | 91.4 | 72.6 | 92.2 | 87.5 | 84.4 | 95.9 | 78.8 | 95.5 | 84.8 | 65.8 | 75.6 | 82.2 | — | — | — | — | |
| RS-CNNinput=2k2019.04 | 86.2 | 84 | 83.5 | 84.8 | 88.8 | 79.6 | 91.2 | 81.1 | 91.6 | 88.4 | 86 | 96 | — | 94.1 | 83.4 | 60.5 | 77.7 | 83.6 | 73.7 | — | — | — | |
| RS-CNN2020.12 | 86.2 | 84 | 83.5 | 84.8 | 88.8 | 79.6 | 91.2 | 81.1 | 91.6 | 88.4 | 86 | 96 | 73.7 | 94.1 | 83.4 | 60.5 | 77.7 | 83.6 | — | — | — | — | |
| RSCNN2023.12 | 86.2 | 84 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| Point-MAETraining protocol=Self-Supervised (Full fine-tuning), #TP (M)=27.062024.08 | 86.1 | 84.19 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| ACTTraining protocol=Self-Supervised (Full fine-tuning), #TP (M)=27.062024.08 | 86.1 | 84.66 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| Point-MAETraining protocol=PEFT baseline, #TP (M)=27.062024.08 | 86.1 | 84.19 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| RECONTraining protocol=PEFT baseline, #TP (M)=27.062024.08 | 86.1 | 84.52 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| Point-MAEBackbone=Point-MAE, Trainable Parameters=27.06M2024.10 | 86.1 | 84.19 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| ACTBackbone=ACT, Trainable Parameters=27.06M2024.10 | 86.1 | 84.66 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| RECONBackbone=RECON, Trainable Parameters=27.06M2024.10 | 86.1 | 84.52 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| PointCNN2020.12 | 86.1 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | 84.6 | — | |
| PointCNN2022.05 | 86.1 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| Point-MAE2022.05 | 86.1 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| PointMLP2022.05 | 86.1 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| PointMLP2022.08 | 86.1 | 84.6 | 83.5 | 83.4 | 87.5 | 80.5 | 90.3 | 78.2 | 92.2 | 88.1 | 82.6 | 96.2 | — | 95.8 | 85.4 | 64.6 | 83.3 | 84.3 | 77.5 | — | — | — | |
| ASSANet2022.11 | 86.1 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| PointMLP2022.11 | 86.1 | 84.6 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| PointCNN2023.12 | 86.1 | 84.6 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| PointASNL2023.12 | 86.1 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| PAConv2023.12 | 86.1 | 84.6 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| PointMLP2023.12 | 86.1 | 84.6 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| Point-MAETraining Strategy=with Self-supervised pretraining, Parameters (M)=27.1, FLOPs (G)=15.52024.10 | 86.1 | 84.2 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| Point-MAEPublication=ECCV'22, TP=27.06 M, Fine-tuning Protocol=Full Fine-Tuning2025.04 | 86.1 | 84.19 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| ACTPublication=ICLR'23, TP=27.06 M, Fine-tuning Protocol=Full Fine-Tuning2025.04 | 86.1 | 84.66 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| RECON (Full-FT)Publication=ICML'23, TP=27.06 M, Fine-tuning Protocol=Full Fine-Tuning2025.04 | 86.1 | 84.52 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| PointMLPInputs=2k, Param (M)=16.8, GFLOPs (G)=6.2, Train Time=47.1 h2026.01 | 86.1 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| Point-MAE2026.03 | 86.1 | 84.2 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| ACT2026.03 | 86.1 | 84.7 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| ReCon2026.03 | 86.1 | 84.5 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| Point-MAE2026.07 | 86.1 | 84.19 | 84.3 | 85 | 88.3 | 80.5 | 91.3 | 78.5 | 92.1 | 87.4 | 86.1 | 96.1 | 75.2 | 94.6 | 84.7 | 63.5 | 77.1 | 82.4 | — | — | — | — | |
| MOSTBackbone=ReCon [38], #P (M)=1.3, #F (G)=4.8, Training Strategy=with Pre-Training (Parameter-Efficient Fine-Tuning)2025.03 | 86 | 84.42 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| PointMambaTraining Strategy=with Self-supervised pretraining, Parameters (M)=17.4, FLOPs (G)=14.32024.10 | 86 | 84.4 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| Mamba3d+MAPTraining Strategy=with Self-supervised pretraining, Parameters (M)=23, FLOPs (G)=11.82024.10 | 86 | 84.5 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| MaskPointPublication=ECCV'22, Fine-tuning Protocol=Full Fine-Tuning2025.04 | 86 | 84.6 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| PAConvInputs=2k2026.01 | 86 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| Full fine-tuningPre-trained model=Uni3D-S2025.02 | 86 | 84.6 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| MaskPoint2026.03 | 86 | 84.6 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| ACT + IDPTBackbone=ACT, Trainable Parameters=5.69M2024.10 | 85.9 | 83.82 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| GBNet2022.05 | 85.9 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| Full fine-tuningPre-trained model=Point-MAE2025.02 | 85.9 | 84.6 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| Point-MAE + PointGSTBackbone=Point-MAE, Trainable Parameters=5.59M2024.10 | 85.8 | 83.81 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| ACT + PointGSTBackbone=ACT, Trainable Parameters=5.59M2024.10 | 85.8 | 84.04 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| RECON + PointGSTBackbone=RECON, Trainable Parameters=5.59M2024.10 | 85.8 | 83.98 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| PointGPT-L + PointGSTBackbone=PointGPT-L, Trainable Parameters=31.85M2024.10 | 85.8 | 83.91 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| SGPN2020.12 | 85.8 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | 82.8 | — | |
| Mamba3d+P-MTraining Strategy=with Self-supervised pretraining (Point-MAE strategy), Parameters (M)=23, FLOPs (G)=11.82024.10 | 85.8 | 84.3 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| APESPublication=CVPR 23, Fine-tuning Protocol=Traditional Supervised Learning Only2025.04 | 85.8 | 83.67 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| APES(global)Inputs=2k, Param (M)=2.0, GFLOPs (G)=15.562026.01 | 85.8 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| PointGSTPre-trained model=Point-MAE2025.02 | 85.8 | 83.8 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| PointGSTPre-trained Encoder=Point-MAE2026.03 | 85.8 | 83.8 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| PointGSTPre-trained Encoder=ReCon2026.03 | 85.8 | 84 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| Gaussian-MAEreproduced=true2026.07 | 85.8 | 83.65 | 84.8 | 84.8 | 90.2 | 80.2 | 90.8 | 73.1 | 91.8 | 87.6 | 84.4 | 95.9 | 76.6 | 95.9 | 85.3 | 61.5 | 73.9 | 81.9 | — | — | — | — | |
| IDPTTraining protocol=PEFT, Backbone=Point-MAE, #TP (M)=5.692024.08 | 85.7 | 83.79 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| DAPTTraining protocol=PEFT, Backbone=Point-MAE, #TP (M)=5.652024.08 | 85.7 | 84.01 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| PPTTraining protocol=PEFT, Backbone=Point-MAE, #TP (M)=5.622024.08 | 85.7 | 84.07 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| IDPTTraining protocol=PEFT, Backbone=RECON, #TP (M)=5.692024.08 | 85.7 | 83.66 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| DAPTTraining protocol=PEFT, Backbone=RECON, #TP (M)=5.652024.08 | 85.7 | 83.87 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| Point-BERT + PointGSTBackbone=Point-BERT, Trainable Parameters=5.58M2024.10 | 85.7 | 83.87 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| Point-MAE + IDPTBackbone=Point-MAE, Trainable Parameters=5.69M2024.10 | 85.7 | 83.79 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| Point-MAE + DAPTBackbone=Point-MAE, Trainable Parameters=5.65M2024.10 | 85.7 | 84.01 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| RECON + IDPTBackbone=RECON, Trainable Parameters=5.69M2024.10 | 85.7 | 83.66 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| RECON + DAPTBackbone=RECON, Trainable Parameters=5.65M2024.10 | 85.7 | 83.87 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| PointConv2020.12 | 85.7 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | 82.8 | — | |
| P2PBackbone=ConvNeXt-Base, Head=SemanticFPN2022.08 | 85.7 | 82.5 | 83.2 | 84.1 | 85.9 | 78 | 91 | 80.2 | 91.7 | 87.2 | 85.4 | 95.4 | — | 93.5 | 79.4 | 57 | 73 | 83.6 | 69.6 | — | — | — | |
| PointConv2023.12 | 85.7 | 82.8 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| MOSTBackbone=Mamba3D [20], #P (M)=1.2, #F (G)=3.9, Training Strategy=with Pre-Training (Parameter-Efficient Fine-Tuning)2025.03 | 85.7 | 83.74 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| Mamba3DTraining Strategy=Supervised Learning Only, Parameters (M)=23, FLOPs (G)=11.82024.10 | 85.7 | 83.7 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| Mamba3d+P-BTraining Strategy=with Self-supervised pretraining (Point-BERT strategy), Parameters (M)=21.9, FLOPs (G)=9.52024.10 | 85.7 | 84.1 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — |