Large Scale Face Recognition In the Wild: Technical Challenges and Research Directions

Face recognition (FR) is the most effective and preferable biometric technique for both verification and identification of humans as compared to iris, voice, retina eye scanning, fingerprint, gait, hand, and ear geometry. FR is the most highlighted and fast-growing research area in computer vision for the past couple of years. It is observed in the literature that most techniques don't perform well because of unconstrained environments like occlusion, noise, position, angle of view, lighting, illumination, ageing, bad picture quality, low resolution, blurriness, and in many cases uncertainty in data. In this paper, a critical review on prior mentioned issues and their proposed solutions to resolve these issues are analyzed and presented through the state of the art techniques which has been proposed in the literature. Apart from this, it has been observed that various new techniques are applied in the form of loss function revamping, adding regularization, using transfer/reinforcement learning, and some new proposed architectures in the literature.