A DEEP LEARNING APPROACH FOR DRIVER DROWSINESS DETECTION USING CNN AND MOBILENETV2

Abstract

Driver drowsiness is a significant factor contributing to road accidents worldwide, making its early detection essential for enhancing road safety. This study presents a deep learning-based approach for detecting driver fatigue using facial image analysis. The proposed system employs a Convolutional Neural Network (CNN) with transfer learning using the MobileNetV2 architecture to automatically extract meaningful features from facial images and classify the driver’s state as either alert or drowsy. To improve model performance and robustness, input images are preprocessed and augmented to handle variations in lighting conditions, facial orientation, and image quality. MobileNetV2 serves as an efficient feature extractor, while additional fully connected layers perform accurate classification. The model effectively identifies visual indicators of drowsiness, such as partially closed eyes and reduced facial activity. Additionally, the system supports real-time monitoring by integrating with OpenCV, enabling continuous analysis of driver behavior during vehicle operation. Experimental results demonstrate that the proposed framework achieves reliable and efficient detection while maintaining low computational complexity due to the lightweight nature of MobileNetV2. Overall, the system provides a practical and scalable solution for real-time driver monitoring and can be integrated into intelligent transportation systems and advanced driver assistance systems to reduce accidents caused by driver fatigue.