Bilkent University
Department of Computer Engineering


Retargeting Human Motion to Non-Humanoid Meshes


İlker Oğuzhan Yaz
MSc Student Student
Computer Engineering Department
Bilkent University

This thesis presents an approach for retargeting human motion-captured animation to non-humanoid meshes. In our consideration, human motion data is generic human animation with no specific requirements, and mesh model is simply collection of vertices, edges and faces with no additional structure (e.g. skeleton). Having those in mind, we try to propose a method that retargets animation of human model to target meshes with as little user intervention as possible. In this way, theoretically, it will be possible to animate any existing mesh using any existing human motion-capture data. Our method consists of a mapping function, and a specific shape preserving deformer. The mapping function uses example poses between human and mesh model while finding correspondent mesh pose for a given human pose. We simplify this process by considering a human pose as a joint set, and a mesh pose as a control point set. Specialized shape preserving deformer then determines resulting mesh pose from control point set that are provided by the mapping function.

Keywords: Animation, mesh deformation, motion retargeting


DATE: 05 March, 2012, Monday @ 16:00