采用动态残差图卷积的3D点云超分辨率
|
|
钟帆,柏正尧
|
3D point cloud super-resolution with dynamic residual graph convolutional networks
|
|
Fan ZHONG,Zheng-yao BAI
|
|
| 表 1 点云超分辨率对比 |
| Tab.1 Performance comparison in point cloud super-resolution |
|
| 方法 | CD | EMD | F-score | NUC | mean | std | | CD | EMD | F-score | NUC | mean | std | | 2× | 4× | | 注：CD、EMD、NUC、mean、std指标均为10−3数量级 | | AR-GCN | − | − | − | − | − | − | | 8.6 | 18.0 | 70.09% | 339.0 | 2.9 | 3.3 | | PU-GAN | 16.0 | 9.0 | 32.17% | 249.0 | 12.0 | 15.0 | 9.7 | 16.0 | 69.75% | 202.0 | 3.0 | 3.1 | | AR-GCN x16sample | 15.0 | 13.0 | 36.98% | 273.0 | 6.7 | 8.2 | 13.0 | 13.0 | 54.05% | 288.0 | 6.6 | 8.0 | | PC-SR | 15.0 | 12.0 | 52.82% | 188.0 | 2.8 | 3.3 | 14.0 | 19.0 | 72.18% | 211.0 | 2.8 | 3.1 | | Meta-PU | 10.0 | 6.3 | 53.20% | 163.0 | 2.6 | 2.9 | 8.7 | 7.8 | 74.05% | 192.0 | 2.6 | 2.7 | | PSR-DRGCN | 9.0 | 6.0 | 70.06% | 154.0 | 2.4 | 2.3 | 8.8 | 7.6 | 70.12% | 183.0 | 2.4 | 2.2 | | | 方法 | CD | EMD | F-score | NUC | mean | std | | CD | EMD | F-score | NUC | mean | std | | 6× | 9× | | AR-GCN | − | − | − | − | − | − | | 8.1 | 22.0 | 74.63% | 344.0 | 3.4 | 4.4 | | PU-GAN | 12.0 | 13.0 | 58.56% | 287.0 | 11.0 | 18.0 | 9.1 | 8.5 | 70.61% | 212.0 | 4.7 | 5.7 | | AR-GCN x16sample | 12.0 | 14.0 | 59.41% | 293.0 | 6.5 | 7.9 | 11 | 14.0 | 62.70% | 298.0 | 6.5 | 7.8 | | PC-SR | 14.0 | 22.0 | 70.02% | 225.0 | 2.7 | 2.9 | 9.3 | 25.0 | 72.92% | 232.0 | 2.6 | 2.8 | | Meta-PU | 8.5 | 14.0 | 72.98% | 267.0 | 2.5 | 3.0 | 8.3 | 16.0 | 73.74% | 274.0 | 3.0 | 3.4 | | PSR-DRGCN | 8.3 | 8.4 | 73.56% | 243.0 | 2.3 | 2.1 | 7.9 | 9.8 | 74.23% | 210.0 | 2.2 | 2.0 |
|
|
|