Indoor scene understanding requires a large amount of 3D data. However, building
such data takes a lot of efforts. In this project, we use scanned RGB-D data to
learn a generative model for these 3D scenes. The generative model can later be
used for data generation, scene understanding and human-assited design process.
Required Skill: basic C++/Python programming, computer graphics background is
The students will learn about deep learning techniques for scene generation and
understanding. They will get training on data processing and how to build a 3D