Classification of demographic attributes from facial images

Abstract

Human facial appearance is strongly influenced by demo-graphical characteristics such as categorical age, ethnicity and gender with each category further partitioned into classes White, Black, East Asian, Male, Fe- male, Child (1- 18), Young (19-36), Middle Age (37-54) and old (55-above). Most subjects share a more similar appearance with their own demographic class than with other demographic class. We evaluate here the accuracy of automatic facial verification for subjects belonging to varying age, ethnicity, and gender categories. For this purpose, we use pre-trained convolutional neural network for feature extraction and show that our method yields an acceptable performance on individual demographics for development of a commercial face recognition engine. We have used IMDB data set for training of our system. We have concluded that results on ethnicity groups white and sub-continent are relatively lower than other ethnicity group’s result. We discuss the results and make suggestions for improving facial image classification across varying demographics, in addition to the development of a system.