Drowsiness Detection for Safer Driving using Deep Learning
1 M V N Harshavardhan Reddy, 2 M V Sai Geethika, 3 Y V Nikhil, 4 K V Sai Siva Kumar,
5 Y V Sai Teja, 6 S Jim Reeves
Abstract
Drowsy driving remains a significant challenge to road safety, contributing to numerous accidents and fatalities worldwide. Fatigue impairs cognitive functions, reduces reaction times, and diminishes situational awareness, increasing the likelihood of collisions. To address this critical issue, this study presents a robust and intelligent approach to drowsiness detection by integrating deep learning and computer vision techniques. Leveraging the powerful capabilities of convolutional neural networks (CNNs), the proposed system is designed to analyze intricate patterns in visual data to identify early signs of driver fatigue. The primary focus is on real-time detection based on eye-blinking patterns, which are well-established indicators of drowsiness. The system utilizes a custom dataset for model training, ensuring adaptability to various lighting conditions, facial structures, and driving environments. To achieve precise monitoring, facial landmarks are employed to track eye and mouth movements. These landmarks enable accurate detection of key behavioral cues, such as eye-blink rates and yawning frequency. By continuously monitoring these visual features, the system can effectively assess the driver's alertness level and provide timely warnings if drowsiness is detected. Experimental results reveal a strong correlation between yawning and prolonged eye closure, both of which are reliable physiological markers of fatigue. A threshold-based classification system is implemented to categorize eye states as either "Open" or "Closed," enhancing the accuracy of drowsiness detection. By establishing predefined thresholds for blink duration and yawning frequency, the system can differentiate between normal blinking patterns and fatigue-induced drowsiness. This research underscores the potential of CNNs and computer vision in developing advanced, real-time drowsiness detection systems. By providing real- time alerts, this system can significantly reduce the risks associated with driver fatigue, ultimately contributing to improved road safety and saving lives. Future advancements may include multi- modal integration with physiological sensors, adaptive learning models, and enhanced real-time processing for even greater accuracy and reliability.
