Learning‐based pose edition for efficient and interactive design

Authoring an appealing animation for a virtual character is challenging task. In computer-aided keyframe artists define the key poses of by manipulating its underlying skeletons. To look plausible, pose must respect many ill-defined constraints, and so resulting realism greatly depends on animator's skill knowledge. Animation software provide tools to help in this matter, relying various algorithms automatically enforce some these constraints. The increasing availability motion capture data has raised interest data-driven approaches design, with potential shifting more task assessing from artist computer, easier access nonexperts. article, we propose such method, neural networks learn constraints data. We describe efficient tool allowing Naïve users intuitively manipulate create animations.