Implementation and Performance Analysis of Face Recognition System

Face recognition system is a computer application for automatically identifying a person from still image or video source. A modern face recognition pipeline consists of four stages: detect face part, face alignment, representation of facial image, and recognition or classification. Though there is a lot of research going on the field of face recognition, achieving high recognition accuracy is still a challenging task for current approaches due to variations in the face images [1]. Recently, deep convolution neural networks set a new trend in the field of face recognition by improving state of art performance [4]. This project addresses deep convolution neural network based face recognition system using OpenFace tool and performance analysis of the system based on pose, illumination variations in the image, and changing size of training dataset. We use histogram of oriented gradients for face detection, aligning faces based on face landmark estimation algorithm and support vector machine for classification. Keywords— Face representation, face detection, face recognition, training, testing.