Point Cloud Registration On 3Dmatch At Least 2
Metrics
Recall ( correspondence RMSE below 0.2)
Results
Performance results of various models on this benchmark
Model Name | Recall ( correspondence RMSE below 0.2) | Paper Title | Repository |
|---|---|---|---|
| OMNet (reported in REGTR) | 35.9 | OMNet: Learning Overlapping Mask for Partial-to-Partial Point Cloud Registration | |
| 3DSN (reported in PREDATOR) | 78.4 | The Perfect Match: 3D Point Cloud Matching with Smoothed Densities | |
| Predator-5k | 89 | PREDATOR: Registration of 3D Point Clouds with Low Overlap | |
| Predator-1k | 90.5 | PREDATOR: Registration of 3D Point Clouds with Low Overlap | |
| REGTR | 92 | REGTR: End-to-end Point Cloud Correspondences with Transformers | |
| D3Feat (reported in PREDATOR) | 81.6 | D3Feat: Joint Learning of Dense Detection and Description of 3D Local Features | |
| PCAM (reported in REGTR) | 85.5 | PCAM: Product of Cross-Attention Matrices for Rigid Registration of Point Clouds | |
| NgeNet | 92.9 | Leveraging Inlier Correspondences Proportion for Point Cloud Registration | |
| FCGF (reported in PREDATOR) | 85.1 | Fully Convolutional Geometric Features | |
| Predator-NR | 62.7 | PREDATOR: Registration of 3D Point Clouds with Low Overlap | |
| DGR (reported in REGTR) | 85.3 | Deep Global Registration |
0 of 11 row(s) selected.