A survey on partially occluded faces

Abstract

Over the past decade, partially occluded face recognition has been an urgent challenge to computer visionaries due to conditions, which appear unconstrained. The main aim of the facial recognition system is to attain the ability to detect partially occluded regions of an individual’s face and authenticating/verifying that face. There are existing neural networks that are proven to be perfect on analysing the patterns for constrained looks but fail to perform in analysing partially occluded faces that are common in the real world. The paper discusses the trainable Deep Learning Neural Network (DLNN) for partially occluded faces by recognizing all the possible faces in the image, either resting, posing or projecting faces and matching them across the trained datasets of DLNN and encoding the identified faces.