A hierarchical face recognition algorithm
Citation
Export citationBouckaert, R. R. (2009). A hierarchical face recognition algorithm. In Z.-H. Zhou and T. Washio (Eds.), Proceedings of the First Asian Conference on Machine Learning, ACML 2009, Nanjing, China, November 2-4, 2009 (pp. 38-50). Berlin, Germany: Springer.
Permanent Research Commons link: https://hdl.handle.net/10289/8015
Abstract
In this paper, we propose a hierarchical method for face recognition where base classifiers are defined to make predictions based on various different principles and classifications are combined into a single prediction. Some features are more relevant to particular face recognition tasks than others. The hierarchical algorithm is flexible in selecting features relevant for the face recognition task at hand. In this paper, we explore various features based on outline recognition, PCA classifiers applied to part of the face and exploitation of symmetry in faces. By combining the predictions of these features we obtain superior performance on benchmark datasets (99.25% accuracy on the ATT dataset) at reduced computation cost compared to full PCA.
Date
2009Publisher
Springer