Statistical 3D Shape Models of Human Faces

Timo Bolkart, Alan Brunton, Augusto Salazar, Stefanie Wuhrer

Statistical 3D shape models of human faces have a variety of applications, such as the generation of realistic synthetic face models or the reconstruction and tracking of detailed 3D face models from input images or point clouds. These faces can then be used in applications, such as image or video editing, tele-presence, or ergonomic design.
For our research, we have computed the following four high-quality statistical 3D shape models of human faces based on the BU-3DFE face database. We make all four statistical models available for non-commercial research purposes. We ask that you respect the conditions of using the models, which are detailed in the readme.pdf files provided with the models.

Template fitting for non-rigid mesh registration

Template fitting used to register the training data of the statistical model described in

A. Brunton, A. Salazar, T. Bolkart, S. Wuhrer
Review of Statistical Shape Spaces for 3D Data with Comparative Analysis for Human Faces
Computer Vision and Image Understanding, 128:1-17, 2014
[DOI] [tech report]

download template fitting code

Statistical models of 3D face shapes of different subjects in neutral expression

Statistical model computed using wavelet-based principal component analysis described in

A. Brunton, A. Salazar, T. Bolkart, S. Wuhrer
Review of Statistical Shape Spaces for 3D Data with Comparative Analysis for Human Faces
Computer Vision and Image Understanding, 128:1-17, 2014
[DOI] [tech report]

download model & fitting code
Statistical model computed using principal component analysis visualized here and described in

A. Brunton, A. Salazar, T. Bolkart, S. Wuhrer
Review of Statistical Shape Spaces for 3D Data with Comparative Analysis for Human Faces
Computer Vision and Image Understanding, 128:1-17, 2014
[DOI] [tech report]

download model & fitting code

Statistical model of 3D face shapes of different subjects in different expressions

Statistical model computed using wavelet-based multilinear analysis described in

A. Brunton, T. Bolkart, S. Wuhrer
Multilinear Wavelets: A Statistical Shape Space for Human Faces
European Conference on Computer Vision, 2014
[DOI] [tech report] [video]
Copyright notice: Springer. The final version of this article is available on the publisher's website.

download model & fitting code
Statistical model computed using multilinear analysis visualized here and described in

T. Bolkart, S. Wuhrer
3D Faces in Motion: Fully Automatic Registration and Statistical Analysis
Computer Vision and Image Understanding, 131:100115, 2015
[DOI]

and

T. Bolkart, S. Wuhrer
Statistical Analysis of 3D Faces in Motion
3D Vision, 2013, pages 103-110
[DOI]

download model & fitting code

Groupwise statistical 3D face model learning methods

Statistical models computed using a groupwise multilinear correspondence optimization described in

T. Bolkart, S. Wuhrer
A Groupwise Multilinear Correspondence Optimization for 3D Faces
International Conference on Computer Vision, 2015
[project page]
Statistical model computed using a robust multinear model learning framework described in

T. Bolkart, S. Wuhrer
A Robust Multilinear Model Learning Framework for 3D Faces
Conference on Computer Vision and Pattern Recognition, 2016
[project page]

Acknowledgments

All face models were computed using the BU-3DFE face database. We wish to thank Lijun Yin for making this database available and for giving us permission to make the statistical models available for non-commerical research purposes. If you use one of our statistical models in your publications, please also reference the following work.
  • L. Yin, X. Wei, Y. Sun, J. Wang, M. Rosato
    A 3D Facial Expression Database For Facial Behavior Research
    International Conference on Automatic Face and Gesture Recognition, 2006, pages 211-216
This work is supported by the Cluster of Excellence on Multimodal Computing and Interaction.