Safedrivevision – A Safe Driving Monitoring...

Safedrivevision – A Safe Driving Monitoring System Using Deep Learning

Author Name : V. Sai Shashank, Thanish Kumar, J. Sai Rahul, P. Vennela, K. Sai Rithick, G. Sai Srinish

ABSTRACT Driver drowsiness detection is a crucial area of research aimed at reducing road accidents caused by fatigued drivers. Traditional methods for detecting drowsiness often rely on physical sensors or behavioural cues, which may be intrusive or inconsistent due to varying environmental conditions. The existing system leveraged a basic eye aspect ratio (EAR) method combined with threshold-based classification, achieving an accuracy of 96.5%. However, this approach struggled with variations in lighting conditions and driver movements, leading to false positives and reduced reliability. Additionally, prolonged driving sessions were not adequately factored into the alert mechanism. The proposed system integrates deep learning with computer vision techniques to enhance detection accuracy and robustness. It utilizes a convolutional neural network (CNN) trained on facial landmarks, combined with dlib's facial recognition capabilities and DeepFace for additional expression analysis. The model achieves a detection accuracy of 96%, significantly improving reliability over traditional methods. Furthermore, the system introduces a driving time tracker, issuing warnings when a driver has been active for more than 10 hours, ensuring long-term fatigue detection. The implementation is designed with an intuitive single-page interface using Tkinter, making it user-friendly and efficient. This app