3D Point Cloud Reconstruction Based on Deformed Network

With development of deep learning, 3D reconstruction has become more and more popular based on it. For complexity and variability of 3D reconstruction object itself, overall reconstruction quality of 3D reconstruction method based on voxels is not high. 3D reconstruction method based on point cloud has better reconstruction effect than the method based on voxel. In this paper, a fully connected point cloud deformation network and a GraphX-based multi-resolution point cloud deformation network is proposed. Experiments show its efficiency and when IoU is used as evaluation metrics, problem of poor quality evaluation in most voxel-based methods can be solved.