Search

It is well documented that there is a direct relationship between student performance in class and attendance. Therefore, recording attendance is an integral part of the educational process. Going through this process, a large majority of class time is wasted collecting and marking attendance manually. This loss of time becomes bigger as class size increases forcing the instructor to lose valuable time calling out names. This paper presents a non-intrusive method using facial recognition to automate the attendance-taking process in a classroom while minimizing the disturbance in the routine that both students and instructors are used to. The approach is using feature extraction CNN model (FEM) and K-Nearest Neighbor (KNN) algorithm to implement face recognition. The proposed system aims to be highly accurate, and easy to introduce into pre-existing campuses without the need of taking too many images of students that are already enrolled while also being scalable to allow large universities to make use of the system in large classes. The complete system achieved an attendance accuracy of 98% given these constraints, using a single camera placed in front of the students. The camera is located at a height of 2.4 meters at a distance of 2 meters from the first row. Allowing it to capture front images of students which were found to be the best performing in terms of detection and recognition accuracy. This accuracy was achieved while only requiring 5 images per student, from 5 different angles which can be easily captured during the enrolment process of the students. [ABSTRACT FROM AUTHOR]