Face Recognition Attendance Management System (FRAMS) Algorithm Using CNN Model

Abstract

This paper discusses the development of the face recognition attendance management system, FRAMS. In the development of FRAMS, there are two important stages which are face detection algorithm development and face recognition algorithm development. In face detection algorithm development, three proposed face detection methods which are Viola-Jones Haar Cascade Classifier, Local Binary Pattern, LBP and Multi-Task Cascaded Convolutional Neural Networks, MTCNN are used to evaluate their performance in terms of face detection accuracy and total detection time required by those methods. From those experiments, it can be known that MTCNN is the best method as it provides a 100% face detection accuracy compared to another two proposed methods although the processing time is longer than another two methods. In face recognition algorithm development, the pre-trained VGG-16 CNN model is used to perform transfer learning by loading the train and validation image datasets. The confusion matrix is plotted out to evaluate the performance of the trained CNN model. The trained VGG-16 CNN model achieved an accuracy of 99%. Finally, a complete FRAMS has been developed and able to recognise the face feature of students with extremely few misclassification mistakes.