Design and Evaluation of a Real-Time Face Recognition System using Convolutional Neural Networks
K B Pranav, J. Manikandan
Abstract
The advent of high speed processors and high resolution cameras has spearheaded the research towards design of face recognition systems for various applications. Face recognition systems use either offline data or real-time input, based on the application. In this paper, design and evaluation of a real-time face recognition system using Convolutional Neural Network (CNN) is proposed. The initial evaluation of the proposed design is carried out using standard AT&T datasets and the same is later extended towards the design of a real-time system. Details about the tuning of CNN parameters to assess and enhance the recognition accuracy of the proposed system are also reported. A systematic approach to tune the parameters is also proposed to enhance the performance of the system. Maximum recognition accuracies of 98.75% and 98.00% are obtained on using the proposed system with standard datasets and real-time inputs respectively.