Open AccessTransition from Holistic to Deep learning Face Recognition Methods
Author(s) -
Dr.R.S. Sabeenian*,
J. Harirajkumar,
Lizzie D’ Cruz
Publication year - 2019
Publication title -
international journal of recent technology and engineering
Language(s) - English
Resource type - Journals
ISSN - 2277-3878
DOI - 10.35940/ijrte.d7974.118419
Subject(s) - computer science , artificial intelligence , facial recognition system , three dimensional face recognition , convolutional neural network , face detection , pattern recognition (psychology) , feature extraction , face (sociological concept) , deep learning , biometrics , feature (linguistics) , access control , computer vision , field (mathematics) , authentication (law) , face recognition grand challenge , object class detection , computer security , social science , linguistics , philosophy , mathematics , sociology , pure mathematics
Face recognition, the fastest growing biometric technology of computer vision, made a breakthrough in the field of security, healthcare, access control and marketing etc. This technology helps in automatically discern and identify the faces for authentication by comparing available digital image of faces. Various algorithms have been developed for enhancing the performance of face recognition system. The face authentication system entails three major steps, face detection, feature extraction and face recognition. This paper provides some of the major milestones of face representation for recognition like holistic learning approach, feature based approach, hybrid approach and deep learning approach. The various techniques under these categories are reviewed. Finally, implemented face recognition using convolution neural network (CNN). In this method, the image is captured through webcam for the dataset preparation. The detection is carried out by CNN cascade, followed by face landmark and face embedding by FaceNet CNN. Recognition of face is performed after training the network. Implemented faces recognition successfully and accurately for smaller dataset.