Components for Face Recognition

We present a method for automatically learning a set of discriminatory facial components for face recognition. The algorithm performs an iterative growing of components starting with small initial components located around preselected points in the face. The direction of growing is determined by the gradient of the cross-validation error of the component classifiers. In experiments we analyze how the shape of the components and their discriminatory power changes across different individuals and views.

• Type: Conference
• Year: 2004
• Where: FGR
• Authors: Bernd Heisele, Takamasa Koshizen
• Keywords: Biometrics, Discriminatory Facial Components, Discriminatory Power Changes, FGR 2004, Small Initial Components