Human Gender Identification Employing Convolution Neural Networks for Veiled Face Images

Abstract

The interest in biometrics and pattern recognition has seen a notable rise over the past few decades. Gender classification through facial analysis can greatly enhance the process of human identification, distinguishing between male and female individuals. Nevertheless, biometric recognition that focuses on specific facial features, such as the eyes, represents a significant and expansive area of research, particularly from security and social viewpoints in Yemen and other Muslim nations. This study seeks to advance the field of gender identification by introducing a novel model that leverages the capabilities of VGG19 and EfficientNetB7. Experimental results obtained from a veiled faces database indicate that the proposed model outperforms its counterparts, achieving an impressive accuracy rate of 99.1%, in contrast to VGG19's 97.3% and EfficientNetB7's 95.5%.