Matt’s research focuses on applying machine learning for automating complex tasks like 3D scene design, playing video games, and image+video editing. Recently he has worked on designing representations for 3D geometry that can be generated by deep networks, and applying generative and adversarial image networks to create new tools for artists.