GENDER CLASSIFICATION THROUGH CONVNET USING GAIT ENERGY IMAGE AND ONE-SHOT LEARNING

Abstract

Many human attributes like finger print, iris are being prominently used for identification

purpose. Acquisition of these features might be problematic if the person is non- cooperative or not available in vicinity. Another problem that might arise is of cameras.

The facial features require high power cameras for clarity and sometimes the angle in which camera is positioned might prove problematic. Another human identification feature that has been gaining popularity is gender. In current scenario facial features are most popularly used for the purpose, but this might face the same problems as stated above. In such cases using gait for the purpose of identification looks promising. In this paper we have used One shot learning with Siamese network for classification of gender using Gait Energy Images. The model achieves 99% accuracy using the CASIA-B database.