Table 2.
Unoriented normal evaluation on PCPNet dataset and SceneNN dataset.
| Method | Source | PCPNet Dataset | SceneNN | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|
| Noise | Density | Clean | Noise | Average | |||||||
| None | 0,12% | 0,60% | 1,20% | Stripes | Gradient | Average | |||||
| PCPNet37 | 45 | 9,64 | 11,51 | 18,27 | 22,84 | 11,73 | 13,46 | 14,58 | 20,86 | 21,40 | 21,13 |
| HSurf-Net43 | 45 | 4,17 | 8,78 | 16,25 | 21,61 | 4,98 | 4,86 | 10,11 | 7,55 | 12,23 | 9,89 |
| GraphFit44 | 35 | 5,21 | 8,96 | 16,12 | 21,71 | 6,30 | 5,86 | 10,69 | n.a. | n.a. | n.a. |
| MSECNet45 | 45 | 3,84 | 8,74 | 16,10 | 21,05 | 4,34 | 4,51 | 9,76 | 6,94 | 11,66 | 9,30 |
| SHS-Net35 | 35 | 3,49 | 8,43 | 15,73 | 21,05 | 4,45 | 4,17 | 9,55 | 7,30 | n.a. | n.a. |
| PCA | * | 12,48 | 12,92 | 18,19 | 26,96 | 13,95 | 13,09 | 16,27 | 16,37 | 16,69 | 16,53 |
Reported are the RMSE results under different noise levels and different sampling strategies to simulate sensor noise. Lower is better. The source column declares the resource for the test results. * Indicates our own results.