论文部分内容阅读
In this talk,I will introduce a series of our recent works on learning deep models for structure-aware 3D shape generation.Complex 3D shapes can usually be summarized with a coarsely de-efined structure which is consistent and robust across variety of observations.However,most e-xisting deep generative models for 3D shapes do not account for shape structure during training.thus often generating implausible and structurally unrealistic shapes.