A Real-Time System for Monitoring Student Drowsiness in the Classroom Using the Deep Learning Model YOLOv8

• Published in: 2024 5th International Conference on Artificial Intelligence and Data Sciences, AiDAS 2024 - Proceedings
• Main Author: Hamidi N.H.H.M.; Abidin N.A.Z.; Aminuddin R.; Sheng C.C.; Samah K.A.F.A.; Nasir S.D.N.M.
• Format: Conference paper
• Language: English
• Published: Institute of Electrical and Electronics Engineers Inc. 2024
• Online Access: https://www.scopus.com/inward/record.uri?eid=2-s2.0-85209638895&doi=10.1109%2fAiDAS63860.2024.10730451&partnerID=40&md5=83601724526af15933997121a7206b6a

Many students experience a decline in focus and cognitive performance during class, leading to drowsiness that can impact their studies. Drowsiness is a change in an individual's psychobiological state caused by various activities, often associated with stress, fatigue, and boredom. Therefore, lecturers must maintain students' attention to ensure effective learning. However, monitoring every student's attention is challenging for lecturers. As a result, this project aims to help lecturers monitor students during classroom lessons using a web-based real-time detection system. The deep learning model-based You Only Look Once Version 8 is utilized for object detection. The dataset consists of drowsy and awake images collected from Kaggle. Data augmentation, including crop, rotation, shear, hue, saturation, brightness, exposure, and blur, is applied to increase the dataset. The model achieves approximately 97% mean-average precision accuracy and 85% testing accuracy. For future work to improve learning during training, consider increasing the number of images and adding features to the system, such as an alert system that triggers an alarm if a student is detected as drowsy three times. © 2024 IEEE.