Development of Face Recognition System by Using Deep Learning and Face-Net Algorithm in the Operations Processes

Abstract

AbstractBiometry is a branch of science that studies the physical and behavioral characteristics of living things. Every human being has its physical characteristics. These features are unique features that distinguish people from others. Biometric systems are the systems established by examining the biometric properties of people to determine their identity. Facial recognition systems, a type of biometric system, using the facial features of individuals, identify snow-bodies, identify information at the distance between the eyes, nose, cheekbones, jawline, jaw and so on and individual. The real purpose here is to confirm for whom did the operation. With the development of technology, many organizations use biometric data as an authentication tool. Within the scope of our study, to check the accuracy of customers who make transactions by identification method during operational transactions. In this study, if the level of accuracy in the face detection system is above a certain rate, no process blocking is performed. In cases where the similarity is low, alert information is sent to the operator and re-authentication with different data is want to be required. This study has been proposed to overcome the difficulties arising from a new convolutional neural network (CNN) framework, matrix completion, and deep learning techniques. FaceNet algorithm, which is one of the deep face recognition algorithms that can be used in the recognition of face images, is the algorithm used in the study. FaceNet is a neural network that learns to map distances from face images to a compact Euclidean space that corresponds to the face similarity measure. In this study, we will explore how to develop a face detection system using the FaceNet algorithm to identify people from photographs. The picture of the customer taken during an operational transaction is converted into a matrix expression using a trained system using this algorithm. If the matrix value obtained is greater than the threshold in the learned system, the definition is considered positive. If the obtained matrix value is less than the threshold value in the learned system, the definition is considered negative. Our data and success rates show the applicability of the study.KeywordsFace recognitionDeep learningConvolutional Neural Network (CNN)FaceNet