Shallow Convolutional Neural Network for Gender Classification Based on Hand

Abstract

Gender recognition based on the hand image is used in computer vision for human-computer communication, hand-based authentication, and identification systems. Beside this, gender recognition may be applied for criminal investigations, visual surveillance, and other legal purposes. The traditional manual methods require a lot of time and are susceptible to variable fluctuations. However, for low amounts of data, the deep-learning models are going to be overfitted. In this regard, this work proposes a shallow convolutional neural network (CNN) with a regularization method. Here, different gender recognition models are built to detect the gender individually from dorsal and palmar hand images. For that, the 11K hand dataset is divided into four labels, i.e., men dorsal side, women dorsal side, men palm side, and women palm side. These data have been pre-processed by resizing and scaling. Furthermore, a model is developed for recognizing gender from the real time data. According to the experimental results, the model developed for the dorsal hand images outperforms the other proposed models and the current state-of-the-art.