Real Time Face Recognition based on Convolution Neural Network

Abstract

: Face recognition is an active research area with many successful traditional and deep learning methods. In our research, we are introducing a Real Time Face Recognition based on Convolution Neural Network (CNN) method for face recognition and to maintain a high accuracy which is running in real time. Face represents multidimensional, complex, meaningful visualization and developing a real time face recognition computational model for face recognition is not easy. In present time, real time face recognition is a challenging problem. The main challenge of face recognition is that how to improve the image recognition performance of the system when it is affected by the variability of non-linear effects. In this paper, 3-layer CNN architecture is proposed for solving the face recognition problem, CNN gives a solution which is capable of solving the facial images related problems that contains varying illumination, poses, occlusions and facial expressions. The test on 115 subject of Essex University database achieve an accuracy of 98.5% which is similar range of performance as the best result of previous works. Meaningfully, our system of real time face recognition based on CNN completes the facial recognition process in less than 0.019 seconds.