Facial Gender Classification — Analysis using Convolutional Neural Networks

Abstract

Automatic gender classification is an important and challenging problem. The challenges are magnified by low resolution of input images and partial occlusion of the face in existing datasets. In recent years, using facial components to conduct gender classification and using deeper convolutional neural networks has both achieved high accuracy and recognition. This analysis paper examines the effect of using deeper convolutional neural networks trained on separate facial components and the results are compared with the state-of-the-art gender classification techniques. We also investigate the effects of network settings and parameters surrounding convolutional neural networks, how they affect the overall classification and provide insights into age-related gender classification. The results show that the proposed technique is promising and performs better with larger crop sizes. Our experiments suggest that the proposed technique can classify gender well from mouth, nose and face (less eyes) only.