Gender Classification beyond Visible Spectrum using Shallow Convolution Neural Network
Nilu R. Salim, Nagamuthu Krishnan Sundarasrinivasa Sankaranarayanan, Umarani Jayaraman · 2021
Gender classification contributes promising position in a spectrum of utilities in digital platforms like digital purchase, content based retrieval, decision making, searching, forensics and demographic studies. Face recognition system is highly used in cyber forensics especially for criminal identification where gender classification reduces the suspects list of searches. Gender is one of the soft biometrics which provide some evidence about the users' identity that could be beneficial. Also, it could be easily collected at the time of enrollment. In order to tackle with the problem of illumination variation, face images are captured in the NIR spectrum. A shallow CNN has been proposed to perform gender classification in the NIR spectrum. The proposed shallow CNN model has given good performance accuracies when tested on challenging NIR datasets namely: IITKh, Oulu CASIA, HITSZ and CBSR.