← ArchivePaper2015
Dyna: A Model of Dynamic Human Shape in Motion
Gerard Pons-Moll, Javier Romero, Naureen Mahmood, Michael J. Black
Data-driven model of soft-tissue dynamics learned from 4D body scans, capturing pose-dependent deformations beyond standard skinning.
Abstract
To look human, digital full-body avatars need to have soft-tissue deformations like those of real people. We learn a model of soft-tissue deformations from examples using a high-resolution 4D capture system and a method that accurately registers a template mesh to sequences of 3D scans. Using over 40,000 scans of ten subjects, we learn how soft-tissue motion causes mesh triangles to deform relative to a base 3D body model. Our Dyna model uses a low-dimensional linear subspace to approximate soft-tissue deformation and relates the subspace coefficients to the changing pose of the body. Dyna uses a second-order auto-regressive model that predicts soft-tissue deformations based on previous deformations, the velocity and acceleration of the body, and the angular velocities and accelerations of the limbs. Dyna also models how deformations vary with a person's body mass index (BMI), producing different deformations for people with different shapes. Dyna realistically represents the dynamics of soft tissue for previously unseen subjects and motions. We provide tools for animators to modify the deformations and apply them to new stylized characters.
How to read this
- Category
- Method: data-driven soft-tissue dynamics model
- Contributions
- Dyna, a model of pose-dependent soft-tissue deformation learned from a large set of 4D body scans registered to a template mesh
- A low-dimensional linear subspace for soft-tissue deformation tied to body pose via a second-order autoregressive model using prior deformations plus body and limb velocity and acceleration
- Modeling of how deformations vary with body mass index, with tools for animators to edit deformations and retarget them to stylized characters
- Context
- Builds directly on the SMPL skinned body model (Loper et al.), adding learned dynamic soft-tissue behavior on top of pose-based skinning.Builds on: SMPL: A Skinned Multi-Person Linear Model
- Correctness
- Learned from many scans of a limited subject set; generalization to unseen subjects/motions is claimed but bounded by training coverage, and the linear subspace plus autoregressive predictor may smooth out fast or extreme dynamics.
- Clarity
- Clearly written SIGGRAPH paper; a first pass conveys the model, a second pass clarifies the autoregressive formulation.
- How to read it
- Focus on how pose kinematics map to subspace coefficients and the role of the second-order autoregressive term; second pass for the BMI conditioning and the registration pipeline.
Builds on
Built upon by
Nothing yet.
Related work
- Real-Time Weighted Pose-Space Deformation on the GPU 2006 / CGF
- Learning Skeletal Articulations with Neural Blend Shapes 2021 / SIGGRAPH
- STAR: Sparse Trained Articulated Human Body Regressor 2020 / Eurographics
- Avatar Reshaping and Automatic Rigging Using a Deformable Model 2015 / MIG
Keywords
This page summarises the entry and links to its original source. The archive never hosts or redistributes the publication itself.Show it in the full archive list →