An Optimized CNN Framework for Real-Time Driver Drowsiness Detection

Driver fatigue is one of the primary causes of serious road accidents around the world, especially during long drives or late-night driving. As the driver becomes drowsy, he/she faces reduced reaction time, decreased concentration, and a lack of ability to make critical decisions for road safety. To overcome this safety issue, this research study proposes a real-time system to identify early signs of drowsiness using computer vision and machine learning algorithms. Our proposed system continuously observes natural facial expressions, such as blink rates, yawning rates, and head position changes, to assess the level of alertness of the driver. By using facial landmark detection and calculating the Eye Aspect Ratio (EAR), our system correctly detects whether the driver's eyes are closed or open. If the driver closes his/her eyes beyond a certain safety limit or if he/she exhibits chronic signs of fatigue, our system automatically triggers an immediate alert to prevent the possibility of accidents. This proposed system is non-intrusive and requires only a standard camera, making it a cost-effective solution to be incorporated into contemporary vehicles or smartphone applications for road safety. The experimental results show that our system works effectively in normal lighting conditions, detecting fatigue signs before the situation becomes out of control.