Automated target selection for DrivenShape

Abstract

DrivenShape is a data-driven deformation that uses pre-computed data (a.k.a. targets) to approximate the effects of a computationally expensive cloth simulation [Kim and Vendrovsky 2008]. Rather than computing a true, accurate solution, DrivenShape produces a quick approximation that satisfies an acceptable margin of error, where error is defined as a difference in appearance or shape.