| dc.description.abstract | Drowsiness while driving is one of the leading causes of traffic accidents, particularly among drivers of four-wheeled vehicles. This study aims to develop a real-time driver drowsiness detection system based on facial image analysis using the Single Shot Multibox Detector (SSD) algorithm, integrated with the MobileNetV3 architecture and a transfer learning approach. The dataset consists of 3,600 facial images of drivers captured under three different lighting conditions: low light, normal light, and high light. Data were collected from public sources and supplemented with self-acquired data, then underwent several preprocessing steps, including face cropping, lighting classification using the HSV method, image augmentation, and object annotation. The best-performing model was trained for 150 epochs with a batch size of 16, using hyperparameters optimized through grid search. The training process yielded a mAP@0.5 of 0.9347, with an accuracy of 95% on the test dataset and 93% in real-time evaluation. The system is capable of classifying driver states into “awake” and “drowsy” based on six visual parameters and issuing automatic warnings when sustained drowsiness is detected. The results demonstrate that the implementation of SSD-MobileNetV3 is effective for real-time, image-based drowsiness detection in real-world driving scenarios. | en_US |
