Learning Inverse Rig Mappings by Nonlinear Regression

We present a framework to design inverse rig-functions, which map the motion of the joints or the surface of the characters to the animation rig. Animators design scenes using the animation rig, which is a framework that is widely adopted in animation
production, which allows the animators to design the character poses and geometry through various control parameters and interfaces. On the other hand, all the state-of-the-art techniques developed in the area of computer graphics control the characters through raw, low level representations such as joint angles and vertex positions, which is becoming a barrier for adopting them for animation production. Our framework fills in this gap by mapping the raw, low level representations to the character rig. Our solution is to use nonlinear regression techniques for mapping the raw character movements to the parameters of the character rig, by using the example animation sequence designed by the animators using the character rig as the training set. With our system, animators have the freedom to attach any motion synthesis algorithms to an arbitrary rigging and animation pipeline, for immediate editing. This greatly improves the productivity of 3D animation, while retaining the flexibility and creativity of artistic input.