← ArchivePaper2009
A Statistical Model of Human Pose and Body Shape
Nils Hasler, Carsten Stoll, Martin Sunkel, Bodo Rosenhahn, Hans-Peter Seidel
Unified statistical model of human pose and body shape learned from 550 laser scans, capturing pose-dependent muscle deformations.
Abstract
Generation and animation of realistic humans is an essential part of many projects in today's media industry. Especially, the games and special effects industry heavily depend on realistic human animation. In this work a unified model that describes both, human pose and body shape is introduced which allows us to accurately model muscle deformations not only as a function of pose but also dependent on the physique of the subject. Coupled with the model's ability to generate arbitrary human body shapes, it severely simplifies the generation of highly realistic character animations. A learning based approach is trained on approximately 550 full body 3D laser scans taken of 114 subjects. Scan registration is performed using a non‐rigid deformation technique. Then, a rotation invariant encoding of the acquired exemplars permits the computation of a statistical model that simultaneously encodes pose and body shape. Finally, morphing or generating meshes according to several constraints simultaneously can be achieved by training semantically meaningful regressors.
How to read this
- Category
- Method / statistical model of human pose and shape
- Contributions
- A unified statistical model that simultaneously encodes human pose and body shape, capturing muscle deformations as a function of both pose and physique
- A learning-based pipeline using non-rigid registration and a rotation-invariant encoding of exemplars to build the model
- Semantically meaningful regressors that morph or generate meshes under several simultaneous constraints
- Context
- Builds on SCAPE (Anguelov et al. 2005), extending data-driven body modeling to jointly couple pose-dependent and shape-dependent deformation in one model.Builds on: SCAPE: Shape Completion and Animation of People
- Correctness
- Trained on roughly 550 full-body laser scans from 114 subjects, so generalization is bounded by that population and by registration quality; the rotation-invariant encoding is the key modeling assumption a reader should keep in mind.
- Clarity
- Accessible motivation with technical core; a first pass conveys the unified pose-plus-shape idea, a second pass for the encoding and regressor training.
- How to read it
- Read first for what unifying pose and shape buys over SCAPE; a second pass on the rotation-invariant encoding and the regressors is worthwhile if you intend to fit or sample bodies.
Builds on
Built upon by
Nothing yet.
Related work
- NeuroSkinning: Automatic Skin Binding for Production Characters with Deep Graph Networks 2019 / SIGGRAPH
- Animation Setup Transfer for 3D Characters 2016 / CGF
- Segmentation-Based Skinning 2017 / CAVW
- NiLBS: Neural Inverse Linear Blend Skinning 2020 / arXiv
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 →