Real Time Recognition of Face Covered with Facemask using CNN for Application in Classroom Attendance System

Abstract

Because each person has a different face identity, facial recognition relies on the distinctness of individual facial traits. However, the recent Covid-19 pandemic has substantially altered daily living, required the usage of facemasks while walking outside, including for educational activities such as attending lectures at schools and universities. Facemasks' widespread use presents a significant problem to facial recognition systems since they obscure a significant area of the face, limiting the availability of important facial clues. In response to this critical issue, this project aims to create a Convolutional Neural Network (CNN) architecture suitable for use in a classroom attendance system. The research process begins with the creation of a large training dataset that includes five different students with four different facial circumstances, including facemask covering. The best CNN design is determined through painstaking testing to enhance accuracy during both the training and validation phases. The empirical findings demonstrated that the finalized CNN architecture has a remarkable validation accuracy of up to 96.2%. The trained system demonstrated outstanding precision in recognizing persons during real-time recognition tests. This experiment highlights the CNN's ability to recognize students with high accuracy even while they are wearing facemasks, indicating its potential in solving the issues provided by the prevailing public health measures.