Output details

Automatic cage construction for retargeted muscle fitting

Originality: Realistic character animation requires the construction of complex anatomical structures that allow subtle shape variation of a character’s outer surface to be displayed in a believable manner. Unfortunately the modelling of muscle structures is still left to the animator making it a tedious, time-consuming and therefore expensive process. Since 2003 we have worked on this via a number of research grants funded by AHRC and GWR.

Our previous work was widely reported internationally (e.g., NewScientist and MIT Technology Review). However because the muscle model was not anatomically correct, it was difficult to generate visually believable skin deformation. Our new retargeting technique can help the animator automatically construct a muscle structure by reusing a pre-existing and physically accurate model. Our technique was the first to achieve automation and anatomical accuracy for muscle modelling in animation production.

Rigour: Our method defines a spatial transfer between the template model and a new model based on the skin surface and the rigging structure. A novel spatial optimization method was proposed to guarantee the retargeted muscle is tightly packed inside a new character. The method was implemented as a plugin for animation software Autodesk Maya and tested by animators with various characters. The experiments and evaluation proved the effectiveness and efficiency of our method.

Significance: Since all stages of the process are completely automatic, no user input is required. In addition the animator is not required to have anatomical knowledge to create realistic accurate musculature models. Not only has this method improved the efficiency of animation production, it has also led to a grant from EU FP7, called AniNex (420,000 Euros, Zhang, Chang), whose aim is to develop next generation computer animation techniques whereby accurate and fast generation of the anatomical structure of a character is essential.