Skinning: Real-Time Shape Deformation
Real-time deformation brings 3D characters to life. Methods developed in recent years abound throughout computer games, film production, medical simulations, and augmented -eality systems. This course examines the motivation and methodology behind state-of-the-art deformation techniques. Traditionally, animators articulate a 3D character by propagating the deformations of an internal skeleton made of rigid bones to the "skin" of the character. The character's geometry might be arbitrarily complex, but so long as the number of bones is small, "skinning" a character is fast. Many linear and non-linear techniques were introduced to blend contributions of individual bones, using either the rest-pose alone or a database of example shapes. The blending weights used to be painted manually, but several automatic techniques are now available. This courses summarizes these recent developments, including application of skinning to arbitrary animations.
The course is organized into four parts:
• Review of the traditional skinning pipeline, including a variety of linear and non-linear transformation blending methods.
• How manual skeleton construction and hand-painting can be avoided with automatic methods
• Discussion of categories of desired qualities and deomonstration of how each can be modeled as a continuous energy or constraint resulting in an optimization discretized using finite-element method and solved efficiently with modern sparse quadratic programming solvers. So far, the direct and automatic methods assume a single pose of the input shape, but artists often want direct control over the resulting poses, leading us to example-based methods that utilize an entire database or training set of shapes, possibly also captured from real individuals.
• Interesting connections between skinning and sparse-matrix decomposition, with applications in compressing arbitrary animations. Summary of various skinning decomposition algorithms that automatically extract skinning weights and transformations from example poses.
University of Houston
University of Pennsylvania
Victoria University of Wellington and Weta Digital