Automated Attendance Tracking in Classrooms Using YOLOv3

Abstract

In contemporary classroom environments, tracking attendance and gauging student comprehension present significant challenges. Traditional methods such as roll calls and scans are prone to errors. Advances in computer vision and machine learning have revolutionized object detection, with notable algorithms like Viola-Jones and HOG detectors laying the groundwork. However, deep learning, particularly convolutional neural networks (CNNs), has significantly enhanced object detection. This study explores the use of the YOLOv3 model, leveraging the COCO dataset, to automate attendance tracking by modifying it to only detect human figures. Results indicate high confidence levels in most detections, suggesting potential for reliable automated attendance tracking. However, the model's efficacy can be further improved by fine-tuning the dataset and addressing image quality issues. This technology can streamline administrative tasks for teachers, allowing more classroom time to be dedicated to lessons. Today, in the current classroom environment, teachers track attendance through roll calls or scanning the classroom to see which students are absent, which both leave large room for error.