An Efficient Human Face Detection Technique Based on CNN with SVM Classifier

Abstract

Face recognition is a growing technology that has been broadly employed in forensics applications such as unlawful person identification, security, and authentication. Computer vision problems now find application in all spheres of the digital world ranging from a normal person’s routine mobile face ID login to institutional facial attendance system to national security-based software to identify criminals. The proposed work examines the performance of face reorganization model using CNN with SVM classifier. Two phases are involved in the creation of the facial recognition system. The first stage involves picking up or extracting facial features, while the second step involves pattern classification. The convolutional neural network (CNN) has made significant strides in FR technology in recent years but most of the existing models considered only one or two parameters but the proposed work computed all important performance parameters, i.e., accuracy, precision, recall, and F-score. Further, the time required to train the model and its prediction time are also computed in this study. The efficiency and dominance of the proposed method are compared with several face detection algorithms, i.e., Eigenface, Fisherface, and Lbph, and results clearly show the supremacy of the proposed approach over the traditional approaches.