Few-shot Classification on ModelNet40 10-way 10-shot
95.2AccuracyReCon
Evaluation results
| Method | Links | |
|---|---|---|
| ReConParadigm=Pre-S-R2026.03 | 95.2 | |
| ReConParadigm=Pre-S2026.03 | 94.4 | |
| Point-MAEParadigm=Pre-S-R2026.03 | 94.2 | |
| Point-PQAEProtocol=FULL, Training Paradigm=Self-Supervised Representation Learning, Input Modality=single-modal information2025.12 | 94.1 | |
| ReConParadigm=Pre-R2026.03 | 94.1 | |
| PointGPT-SEvaluation Protocol=Full Fine-tuning2026.05 | 94.1 | |
| Point-FEMAEProtocol=FULL, Training Paradigm=Self-Supervised Representation Learning, Input Modality=single-modal information2025.12 | 94 | |
| PointLAMAProtocol=FULL, Training Paradigm=Self-Supervised Representation Learning, Input Modality=single-modal information2025.12 | 94 | |
| CSConProtocol=FULL, Training Paradigm=Self-Supervised Representation Learning, Input Modality=single-modal information2025.12 | 93.6 | |
| ReConEvaluation Protocol=MLP-LINEAR2025.12 | 93.6 | |
| ReConEvaluation Protocol=MLP-32025.12 | 93.6 | |
| PCP-MAEProtocol=FULL, Training Paradigm=Self-Supervised Representation Learning, Input Modality=single-modal information2025.12 | 93.5 | |
| CSConEvaluation Protocol=MLP-32025.12 | 93.4 | |
| PointMamba + MantisEvaluation Protocol=Parameter-Efficient Fine-Tuning2026.05 | 93.4 | |
| ACTLearning Protocol=Full Fine-tuning2023.04 | 93.3 | |
| ACTvoting=false2023.09 | 93.3 | |
| Point-RAEvoting=false2023.09 | 93.3 | |
| ACTPublication=ICLR'23, Fine-tuning Protocol=Full Fine-Tuning2025.04 | 93.3 | |
| RECON (Full-FT)Publication=ICML'23, Fine-tuning Protocol=Full Fine-Tuning2025.04 | 93.3 | |
| DAP-MAEProtocol=FULL, Training Paradigm=Self-Supervised Representation Learning, Input Modality=single-modal information2025.12 | 93.3 | |
| ACTProtocol=FULL, Training Paradigm=Self-Supervised Representation Learning, Input Modality=cross-modal information and teacher models2025.12 | 93.3 | |
| ReConProtocol=FULL, Training Paradigm=Self-Supervised Representation Learning, Input Modality=cross-modal information and teacher models2025.12 | 93.3 | |
| HyperPointParadigm=Pre-S2026.03 | 93.3 | |
| HyperPointParadigm=Pre-S-R2026.03 | 93.3 | |
| ReConEvaluation Protocol=Full Fine-tuning2026.05 | 93.3 | |
| ACTEvaluation Protocol=Full Fine-tuning2026.05 | 93.3 | |
| TAPProtocol=FULL, Training Paradigm=Self-Supervised Representation Learning, Input Modality=cross-modal information and teacher models2025.12 | 93.1 | |
| PointMambaEvaluation Protocol=Parameter-Efficient Fine-Tuning2026.05 | 93 | |
| PCP-MAEEvaluation Protocol=MLP-32025.12 | 92.9 | |
| Point-MAE w/ IDPTLearning Protocol=IDPT2023.04 | 92.8 | |
| RECON+ IDPTPublication=ICCV'23, Fine-tuning Protocol=Efficient Fine-Tuning2025.04 | 92.8 | |
| ZigzagPointMamba + MantisEvaluation Protocol=Parameter-Efficient Fine-Tuning2026.05 | 92.8 | |
| RECON+ PointLoRAPublication=Ours, Fine-tuning Protocol=Efficient Fine-Tuning2025.04 | 92.7 | |
| Point-CMAEProtocol=FULL, Training Paradigm=Self-Supervised Representation Learning, Input Modality=single-modal information2025.12 | 92.7 | |
| Point-MAELearning Protocol=Full Fine-tuning2023.04 | 92.6 | |
| Point-MAEvoting=false2023.09 | 92.6 | |
| I2P-MAEvoting=false2023.09 | 92.6 | |
| Point-MAELearning Mode=Self-supervised Pre-training, Voting=false2024.04 | 92.6 | |
| Point-MAEPublication=ECCV'22, Fine-tuning Protocol=Full Fine-Tuning2025.04 | 92.6 | |
| Point-MAEProtocol=FULL, Training Paradigm=Self-Supervised Representation Learning, Input Modality=single-modal information2025.12 | 92.6 | |
| PointGPTProtocol=FULL, Training Paradigm=Self-Supervised Representation Learning, Input Modality=single-modal information2025.12 | 92.6 | |
| Joint-MAEProtocol=FULL, Training Paradigm=Self-Supervised Representation Learning, Input Modality=cross-modal information and teacher models2025.12 | 92.6 | |
| I2P-MAEProtocol=FULL, Training Paradigm=Self-Supervised Representation Learning, Input Modality=cross-modal information and teacher models2025.12 | 92.6 | |
| Point-MAEParadigm=Pre-S2026.03 | 92.6 | |
| HyperPointParadigm=Pre-R2026.03 | 92.6 | |
| Point-MAEEvaluation Protocol=Full Fine-tuning2026.05 | 92.6 | |
| ACT w/ IDPTLearning Protocol=IDPT2023.04 | 92.4 | |
| MAMBA3D+P-MLearning Mode=Self-supervised Pre-training, Pre-training Strategy=Point-MAE, Voting=false2024.04 | 92.4 | |
| Point-FEMAEEvaluation Protocol=MLP-LINEAR2025.12 | 92.4 | |
| ACTEvaluation Protocol=MLP-32025.12 | 92.4 | |
| Point-FEMAEEvaluation Protocol=MLP-32025.12 | 92.4 | |
| Mamba3DEvaluation Protocol=Parameter-Efficient Fine-Tuning2026.05 | 92.4 | |
| Point-M2AELearning Protocol=Full Fine-tuning2023.04 | 92.3 | |
| Point-M2AEvoting=false2023.09 | 92.3 | |
| Point-M2AEPublication=NeurIPS'22, Fine-tuning Protocol=Full Fine-Tuning2025.04 | 92.3 | |
| Point-M2AEProtocol=FULL, Training Paradigm=Self-Supervised Representation Learning, Input Modality=single-modal information2025.12 | 92.3 | |
| Point-MAEParadigm=Pre-R2026.03 | 92.3 | |
| RECON+ PPTPublication=arXiv'24, Fine-tuning Protocol=Efficient Fine-Tuning2025.04 | 92.2 | |
| CSConEvaluation Protocol=MLP-LINEAR2025.12 | 92.2 | |
| Point-PQAEEvaluation Protocol=MLP-32025.12 | 92 | |
| Point-BERT w/ IDPTLearning Protocol=IDPT2023.04 | 91.9 | |
| RECON+ DAPTPublication=CVPR'24, Fine-tuning Protocol=Efficient Fine-Tuning2025.04 | 91.9 | |
| Mamba3D + MantisEvaluation Protocol=Parameter-Efficient Fine-Tuning2026.05 | 91.7 | |
| MaskPointLearning Protocol=Full Fine-tuning2023.04 | 91.4 | |
| MaskPointvoting=false2023.09 | 91.4 | |
| PointMambaLearning Mode=Self-supervised Pre-training, Voting=false2024.04 | 91.4 | |
| MaskPointLearning Mode=Self-supervised Pre-training, Voting=false2024.04 | 91.4 | |
| MaskPointPublication=ECCV'22, Fine-tuning Protocol=Full Fine-Tuning2025.04 | 91.4 | |
| MaskPointProtocol=FULL, Training Paradigm=Self-Supervised Representation Learning, Input Modality=single-modal information2025.12 | 91.4 | |
| MaskPointEvaluation Protocol=Full Fine-tuning2026.05 | 91.4 | |
| MAMBA3D+P-BLearning Mode=Self-supervised Pre-training, Pre-training Strategy=Point-BERT, Voting=false2024.04 | 91.3 | |
| Point-CMAEEvaluation Protocol=MLP-32025.12 | 91.3 | |
| EPCLLearning Protocol=Full Fine-tuning2023.04 | 91.1 | |
| Point-BERTLearning Protocol=Full Fine-tuning2023.04 | 91 | |
| Point-BERTvoting=false2023.09 | 91 | |
| Point-BERTLearning Mode=Self-supervised Pre-training, Voting=false2024.04 | 91 | |
| Point-BERTPublication=CVPR'22, Fine-tuning Protocol=Full Fine-Tuning2025.04 | 91 | |
| Point-BERTProtocol=FULL, Training Paradigm=Self-Supervised Representation Learning, Input Modality=single-modal information2025.12 | 91 | |
| Point-BERTEvaluation Protocol=Full Fine-tuning2026.05 | 91 | |
| Point-MAEEvaluation Protocol=MLP-32025.12 | 90.6 | |
| ZigzagPointMambaEvaluation Protocol=Parameter-Efficient Fine-Tuning2026.05 | 90 | |
| Transformer-OcCoLearning Protocol=Full Fine-tuning2023.04 | 89.4 | |
| OcCoLearning Mode=Self-supervised Pre-training, Voting=false2024.04 | 89.4 | |
| OcCoPublication=ICCV'21, Fine-tuning Protocol=Full Fine-Tuning2025.04 | 89.4 | |
| PCP-MAEEvaluation Protocol=MLP-LINEAR2025.12 | 89.3 | |
| Point-CMAEEvaluation Protocol=MLP-LINEAR2025.12 | 89.2 | |
| Point-PQAEEvaluation Protocol=MLP-LINEAR2025.12 | 89 | |
| MAMBA3DLearning Mode=Supervised Learning, Voting=false2024.04 | 88.1 | |
| Transformervoting=false2023.09 | 84.6 | |
| TransformerLearning Mode=Supervised Learning, Voting=false2024.04 | 84.6 | |
| ACTEvaluation Protocol=MLP-LINEAR2025.12 | 84.5 | |
| Point-MAEParadigm=Pre-R (10%)2026.03 | 83.6 | |
| Point-MAEEvaluation Protocol=MLP-LINEAR2025.12 | 83.5 | |
| DGCNN-OcCoLearning Protocol=Full Fine-tuning2023.04 | 82.9 | |
| DGCNN+OcCovoting=false2023.09 | 82.9 | |
| DGCNN+OcCoLearning Mode=Self-supervised Pre-training, Pre-training Strategy=OcCo, Voting=false2024.04 | 82.9 | |
| OcCoProtocol=FULL, Training Paradigm=Supervised Learning Only2025.12 | 82.9 | |
| NPNetMethod Category=Non-Parametric2026.03 | 82.5 | |
| Point-GNMethod Category=Non-Parametric2026.03 | 81.6 | |
| SLNet-MMethod Category=Parametric2026.03 | 80 |