Point Cloud Registration on ScanObjectNN Segmented Objects (test)
0.11MAE RFGR
Evaluation results
| Method | Links | |||||||
|---|---|---|---|---|---|---|---|---|
| FGRInput points=2048, Use of normals=true2021.03 | 0.11 | 0.001 | 0.19 | 0.001 | 99.7 | 0.16 | 131 | |
| ReAgent ILPre-trained on=ModelNet40, Input points=2048, Training mode=Imitation Learning2021.03 | 0.77 | 0.006 | 1.33 | 0.012 | 95.7 | 0.3 | 21 | |
| PointNetLKPre-trained on=ModelNet40, Input points=20482021.03 | 0.9 | 0.01 | 1.74 | 0.02 | 92.5 | 1.09 | 45 | |
| ReAgent IL+RLPre-trained on=ModelNet40, Input points=2048, Training mode=Imitation Learning + Reinforcement Learning2021.03 | 0.93 | 0.007 | 1.66 | 0.014 | 95.4 | 0.34 | 21 | |
| ICPInput points=20482021.03 | 5.34 | 0.036 | 10.47 | 0.076 | 88.1 | 2.99 | 19 | |
| DCP-v2Pre-trained on=ModelNet40, Input points=20482021.03 | 7.42 | 0.05 | 14.93 | 0.102 | 72.4 | 4.93 | 54 |