Design of Implementation of Deep Learning Based Face Recognition System

Abstract

This study discusses the face recognition system. In general, the face recognition system is divided into two stages, namelynamely system face detection, which is a pre-processing step followed by a face recognition system. This step will be done quickly by humans but takes a long time for computers. This human ability is what researchers have wanted to emulate in recent years as biometric technology in computer vision to create face recognition models on computers. Deep learning is in the spotlight in the development of machine learning, the reason being that deep learning has achieved extraordinary results in computer vision. Based on this, the author has the idea to create a face recognition system by implementing deep learning using the CNN method and implementing it in the openFace library. The CNN method is still superior and widely used because it has good accuracy. The initial process is taking pictures of the face to be photographed as a data set. From the dataset, face preprocessing will be carried out, namely to extract the face vector features into 128-d and classify the face vector. The contribution of this research is the addition of features to improve the accuracy of the face recognition system using the CNN method. The results of this study obtained a precision value of 98.4%, a recall of 98% and an accuracy of 99.84%.