An Experimental Study on Automatic Face Gender Classification

Abstract

This paper presents an experimental study on automatic face gender classification by building a system that mainly consists of four parts, face detection, face alignment, texture normalization and gender classification. Comparative study on the effects of different texture normalization methods including two kinds of affine mapping and one Delaunay triangulation based warping as preprocesses for gender classification by SVM, LDA and Real Adaboost respectively is reported through experiments on very large sets of snapshot images.