Real-time drowsiness detection using mediapipe and machine learning with mobile systems

This thesis presents the development of a real-time driver drowsiness detection system using mobile phone cameras, integrating Mediapipe and machine learning techniques. Drowsy driving is one of the major causes of traffic accidents and poses a significant risk to road safety. Therefore, continuous monitoring of a driver's alertness level and early detection of drowsiness are of critical importance for accident prevention. In the proposed system, facial expressions and eye movements of drivers are analyzed through the mobile phone camera. The Google Mediapipe library is utilized for facial landmark detection, and blendshape features are extracted to numerically represent subtle facial expressions and movements. These blendshape vectors capture behavioral changes associated with drowsiness, such as eyelid closure, decreased mouth activity, and relaxed facial muscles. The extracted features are processed using machine learning algorithms to classify the driver's state as either "drowsy" or "alert." Balanced datasets were used during model training and testing, and the system's performance was evaluated based on accuracy, sensitivity, and specificity metrics. This study introduces a mobile-based, low-cost, real-time drowsiness detection system with minimal hardware requirements. By leveraging blendshape-based facial analysis, the proposed method achieves high accuracy in identifying drowsiness and offers a practical and accessible solution that can contribute significantly to reducing traffic accidents caused by driver fatigue.