Driver fatigue detection in video images using advanced deep learning techniques

نوع: Type: Thesis

مقطع: Segment: masters

عنوان: Title: Driver fatigue detection in video images using advanced deep learning techniques

ارائه دهنده: Provider: mahdi safari pour

اساتید راهنما: Supervisors: Dr. Hassan Khotanlou

اساتید مشاور: Advisory Professors:

اساتید ممتحن یا داور: Examining professors or referees: Dr. moharram mansourizadeh - Dr. reza mohammadi

زمان و تاریخ ارائه: Time and date of presentation: 2025

مکان ارائه: Place of presentation: seminar room

چکیده: Abstract: Road accidents are among the leading causes of mortality worldwide, with a significant portion attributed to driver fatigue and drowsiness. Driver fatigue reduces concentration, slows reaction times, and increases the likelihood of decision-making errors, thereby substantially raising the risk of traffic accidents. Consequently, the development of intelligent driver fatigue detection systems capable of real-time monitoring and timely alerts holds critical importance. In this research, a hybrid multi-branch fusion approach is proposed for driver fatigue detection. By combining facial visual and structural information, the method employs a Vision Transformer (ViT) for image analysis and Graph Neural Networks (GNNs) for analyzing facial structural features. The outputs of these two branches are fused, enabling the model to accurately classify the driver’s state into “fatigue” and “alertness.” The proposed model was trained and evaluated on the standard Drowsiness Prediction and YawDD Dataset, and the results demonstrated that integrating the strengths of image-based and graph-based models through fusion significantly improves accuracy and enhances robustness against variations in lighting conditions compared to single-model approaches. This study can serve as a valuable step toward the development of intelligent driver monitoring systems and the reduction of fatigue-related accidents.