3D Morphable Models are valuable tools in face perception research. Their construction requires vertex-to-vertex correspondence for all face scans in a database. At ZIB, we are developing face matching methods to establish highly accurate dense correspondence in fully automatic fashion. The goal of this project is to improve face matching by training a convolutional neural network to correspondingly label the vertices of facial meshes.

Please see the attachment for further details.



Visualization Techniques for 3D Image Data
3D Convolutional Neural Networks for Classification of Diseases
Evaluation and Comparison of Linear and Non-linear Approaches for the Analysis and Synthesis of Facial Motion