KT-Net: Knowledge Transfer for Unpaired 3D Shape Completion

Unpaired 3D object completion aims to predict a complete shape from an incomplete input without knowing the correspondence between and shapes. In this paper, we propose novel KTNet solve task new perspective of knowledge transfer. elaborates teacher-assistant-student network establish multiple transfer processes. Specifically, teacher takes as learns shape. The student one restores corresponding And assistant modules not only help student, but also judge learning effect network. As result, makes use more comprehensive understanding geometric shapes in transfer, which enables detailed inference for generating high-quality We conduct experiments on several datasets, results show that our method outperforms previous methods unpaired point cloud by large margin. Code is available at https://github.com/a4152684/KT-Net.