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 2 参数与推理时间对比结果

Tab.2 Performance comparison in parameters and inference time

网络 D/(10−2) k t/ms mean std Pointnet++ 3.13 3.74 1 480 25.30 PU-net 0.46 0.55 777 10.40 AR-GCN 0.29 0.33 785 15.60 PU-GAN 0.30 0.31 684 14.30 DGCNN 2.90 3.20 1 842 41.60 PC-SR 2.64 2.90 844 21.30 Meta-PU 0.29 0.27 2 756 90.20 PSR-DRGCN 0.24 0.22 1 282 12.30