Development of Deep Learning-Based Facial Recognition System

Abstract

Recently, developed algorithms in the face recognition field that are based on deep learning technology have made significant progress. However, face recognition under unconstrained scenarios that is where illumination, image resolution, background clutter, facial pose, expression, occlusion and other factors are not controlled, still under heavy research. In this paper, we explore the problem of identifying a person of interest under this unconstrained conditions. To this end, we make the following contributions: firstly, we have implemented a Convolution Neural Network model (CNN) based on VGG16 architecture, using a fast open framework for deep learning called Keras. Then, a serie of experiments is conducted on the Labeled Faces in the Wild benchmark dataset (LFW), demonstrating that the proposed approach achieved state-of-the-art results.