The role of global and feature based information in gender classification of faces: a comparison of human performance and computational models.

This paper describes a global and feature based representation of face images. We use dimensionality reduction techniques and a support vector machine classifier and show this method performs better than more traditional approaches. We present results of human performance on gender classification tasks and evaluate how the different techniques compare with their performance. The results support the psychological plausibility of the global and feature based representation. This aspect of the work was carried out in collaboration with practitioners in Psychiatry at a local hospital. The paper was selected for publication as an expanded version of a paper presented at ICONIP2004.