End-to-end automated student attendance recording system using surveillance camera

The practice of recording and monitoring student attendance is a fundamental action in various contexts, especially in the school environment. However, due to its manual process, it often consumes a significant part of class time. This paper presents an end-to-end automated student attendance recording system utilizing classroom surveillance cameras. The system used advanced technologies such as computer vision, face recognition, and Convolutional Neural Network (CNN) techniques to streamline the attendance process and enhance the time distribution in the classroom. The study evaluates the performance of five class sessions from a public dataset and three class sessions collected by us at the school studied in this paper through different scenarios with varying image quality and student positioning. Our results highlight the superiority of ResNet29 in detecting and recognizing students, especially in lower-resolution images. Compared to Facenet512, Facenet, and ArcFace models, in terms of final frequency marking accuracy, the model showed superior metric results by having an increase of 50% of accuracy compared to the others, reaching 80% accuracy, as well as demonstrating being superior in precision, recall, F1-score, and AUC-ROC metrics. The system's deployment in a school setting has shown promising results, prompting plans for expansion to additional classrooms. The lightweight and non-intrusive nature of the system aligns with the concept of Next-Generation Smart Classrooms, emphasizing its potential to revolutionize attendance management in educational institutions.