3D Object Classification on ModelNet40 50 (test)
86.3Top-1 AccuracyTAMM
Evaluation results
| Method | Links | |||
|---|---|---|---|---|
| TAMMPre-Training Dataset=Ensembled (no LVIS), Model Architecture=Point-BERT, Zero-shot Protocol=true2024.02 | 86.3 | 96.6 | 98.1 | |
| TAMMPre-Training Dataset=Ensembled (no LVIS), Model Architecture=SparseConv, Zero-shot Protocol=true2024.02 | 85.7 | 96.8 | 98.3 | |
| TAMMPre-Training Dataset=Ensembled, Model Architecture=SparseConv, Zero-shot Protocol=true2024.02 | 85.4 | 96.4 | 98.1 | |
| OpenShapePre-Training Dataset=Ensembled (no LVIS), Model Architecture=Point-BERT, Zero-shot Protocol=true2024.02 | 85.3 | 96.2 | 97.4 | |
| TAMMPre-Training Dataset=Ensembled, Model Architecture=Point-BERT, Zero-shot Protocol=true2024.02 | 85 | 96.6 | 98.1 | |
| OpenShapePre-Training Dataset=Ensembled, Model Architecture=Point-BERT, Zero-shot Protocol=true2024.02 | 84.4 | 96.5 | 98 | |
| OpenShapePre-Training Dataset=Ensembled, Model Architecture=SparseConv, Zero-shot Protocol=true2024.02 | 83.4 | 95.6 | 97.8 | |
| OpenShapePre-Training Dataset=Ensembled (no LVIS), Model Architecture=SparseConv, Zero-shot Protocol=true2024.02 | 82.6 | 95 | 97.5 | |
| ULIPPre-Training Dataset=Ensembled, Model Architecture=Point-BERT, Zero-shot Protocol=true2024.02 | 75.1 | 88.1 | 93.2 | |
| TAMMPre-Training Dataset=ShapeNet, Model Architecture=SparseConv, Zero-shot Protocol=true2024.02 | 74.6 | 88.2 | 94 | |
| TAMMPre-Training Dataset=ShapeNet, Model Architecture=Point-BERT, Zero-shot Protocol=true2024.02 | 73.1 | 88.5 | 91.9 | |
| OpenShapePre-Training Dataset=ShapeNet, Model Architecture=SparseConv, Zero-shot Protocol=true2024.02 | 72.9 | 87.2 | 93 | |
| ULIPPre-Training Dataset=Ensembled (no LVIS), Model Architecture=Point-BERT, Zero-shot Protocol=true2024.02 | 71.4 | 84.4 | 89.2 | |
| OpenShapePre-Training Dataset=ShapeNet, Model Architecture=Point-BERT, Zero-shot Protocol=true2024.02 | 70.3 | 86.9 | 91.3 | |
| PointCLIP v2Pre-Training Dataset=None, Model Architecture=PointCLIP v2, Zero-shot Protocol=true2024.02 | 63.6 | 77.9 | 85 | |
| ReConPre-Training Dataset=None, Model Architecture=Transformer, Zero-shot Protocol=true2024.02 | 61.2 | 73.9 | 78.1 | |
| ULIPPre-Training Dataset=ShapeNet, Model Architecture=Point-BERT, Zero-shot Protocol=true2024.02 | 60.4 | 79 | 84.4 | |
| CLIP2PointPre-Training Dataset=None, Model Architecture=ViT-B/32, Zero-shot Protocol=true2024.02 | 49.5 | 71.3 | 81.2 | |
| PointCLIPPre-Training Dataset=None, Model Architecture=PointCLIP, Zero-shot Protocol=true2024.02 | 19.3 | 28.6 | 34.8 |