PhD Defense: Machine Learning for Non-photorealistic Illustration, Animation, and 3D Characters

As anime-style content becomes more popular on the global stage, we ask whether new vision/graphics techniques could contribute to the artform. However, the highly-expressive and non-photorealistic nature of anime poses additional challenges not addressed by standard ML models, and much of the existing work in the domain does not align with real artist workflows. In this dissertation, we will present work building foundational 2D/3D infrastructure in the anime domain (including pose estimation, video frame interpolation, and 3D character reconstruction), as well as new tools for professional 2D animators that leverage novel vision/graphics techniques to assist drawing.