Unconstrained Face Recognition System using Deep Neural Network: A Review

Abstract

The human face is one of the easiest bio-metric modality that does not require any special imaging techniques or subject cooperation. This benefits a wide range of applications, from automatic image tagging in social media structures to critical applications such as border control and surveillance forensics. Different other modalities of bio-metrics are also accurate and sophisticated, but they can hardly replace the need for facial bio-metrics. The success of face recognition technology is limited to unconstrained facial image acquisition scenarios. In the unconstrained scenarios acquiring of face images is not well controlled and subjects may be uncooperative, hence, degrading the performance of face recognition systems. Such confounding factors include variations in facial pose and expression, nonuniform illumination, occlusion and low image resolution. This paper details the state-of-the-art facial recognition techniques in uncontrolled environment and compare on the basis of their accuracy.