Data-Driven Physics for Human Soft Tissue Animation




Data driven models of human pose and soft-tissue deformations can produce very realistic results. However, they only model the visible surface of the human body, and thus cannot create skin deformation due to interactions with the environment. Physical simulation generalizes to external forces but its parameters are difficult to control. In this paper we present a layered volumetric human body model learned from data. Our model is composed of data-driven inner layer and a physics-based external layer. The inner layer is driven with a volumetric statistical body model (VSMPL). The soft tissue layer consists of a tetrahedral mesh that is driven using FEM. The combination of both layers creates coherent and realistic full-body avatars that can be animated and generalize to external forces. Model parameters, namely the segmentation of the body into layers and the soft tissue elasticity are learned directly from 4D registrations of humans exhibiting soft tissue deformations, and the learned parameters can faithfully reproduce the 4D registrations. The resulting avatars produce realistic results for held out sequences and react to external forces. Moreover, the model allows to retarget physical properties from an avatar to another one as they all share the same topology.

Author(s): Meekyoung Kim and Gerard Pons-Moll and Sergi Pujades and Sungbae Bang and Jinwwok Kim and Michael Black and Sung-Hee Lee
Journal: ACM Transactions on Graphics, (Proc. SIGGRAPH) [conditionally accepted]
Year: 2017

Department(s): Perceiving Systems
Bibtex Type: Article (article)
Paper Type: Journal


  title = {Data-Driven Physics for Human Soft Tissue Animation},
  author = {Kim, Meekyoung and Pons-Moll, Gerard and Pujades, Sergi and Bang, Sungbae and Kim, Jinwwok and Black, Michael and Lee, Sung-Hee},
  journal = {ACM Transactions on Graphics, (Proc. SIGGRAPH) [conditionally accepted]},
  year = {2017}