Subspace methods such as PCA, LDA, ICA have become a standard tool to perform visual learning and recognition. In this paper we propose representational oriented component analysis (ROCA), an extension of OCA, to perform face recognition when just one sample per training class is available. Several novelties are introduced in order to improve generalization and efficiency: (1) combining several OCA classifiers based on different image representations of the unique training sample is shown to greatly improve the recognition performance. (2) To improve generalization and to account for small misregistration effect, a learned subspace is added to constrain the OCA solution, (3) a stable/efficient generalized eigenvector algorithm that solves the small size sample problem and avoids overfitting. Preliminary experiments in the FRGC Ver 1.0 dataset show that ROCA outperforms existing linear techniques (PCA, OCA) and some commercial systems.
Representational oriented component analysis (ROCA) for face recognition with one sample image per training class
01.01.2005
1215689 byte
Aufsatz (Konferenz)
Elektronische Ressource
Englisch
Component-based LDA Method for Face Recognition with One Training Sample
British Library Conference Proceedings | 2003
|Managing the Roca electrification
Tema Archiv | 1983
|Roca electrification out to tender
British Library Online Contents | 2008