Date: October 2017
Source: 2017 IEEE International Conference on Computer Vision Workshops (ICCVW), pp. 786-794. DOI: 10.1109/ICCVW.2017.98.
Research Summary: Building directly on the Liverpool-York Head Model, researchers at the University of York and Alder Hey Children’s Hospital developed the first 3D morphable model to separate symmetric from asymmetric craniofacial shape variation. The team’s pipeline symmetrises each 3D head scan using a Laplace-Beltrami regularized method, aligns the dataset using a new symmetry-aware variant of Generalised Procrustes Analysis (GPA), then models symmetric and asymmetric shape variation as two distinct components using Principal Component Analysis (PCA). Built from 1,212 subjects in the real-world 3dMD Headspace dataset, the resulting model outperformed standard PCA models on both compactness and generalisation error, and performed better at half-head completion, reconstructing a full head from just one side of the symmetry plane. Using a linear SVM with 10-fold cross-validation, the team also found that combining symmetric and asymmetric shape parameters improved age-group classification accuracy over either alone.
3dMD’s Role: Built from the same 1,212-subject Headspace dataset as the founding LYHM paper, captured on a 3dMDhead5 system.
Article: Symmetry-factored Statistical Modelling of Craniofacial Shape.
Authors: Hang Dai, William A. P. Smith, Nick Pears. Department of Computer Science, University of York. Christian Duncan, Alder Hey Hospital, Liverpool.