Scholarlink

<p>Guaranteeing safe guiding through dynamic environments is a critical mission for autonomous vehicles (AVs). Driver drowsiness is a severe problem that endangers drivers and road safety. Potential accidents can occur because drivers frequently fail to identify indicators of fatigue. This paper proposes a generative adversarial network (GAN) drowsiness detection model using YOLO (GAN-DD-YOLO) for early drowsiness detection. The GAN generates synthetic data that resembles real-world images of drowsy states, such as closed eyes or yawning. The YOLO model was trained on synthetic and real data to increase the model's performance in identifying indications of sleepiness. The proposed model treats each eye separately, which enhances the detection process due to the capability of covering more angles. The proposed model implemented on a Raspberry Pi linked to several sensors and the Internet of things (IoT) network alerts drivers when drowsiness is detected. Whenever the driver experiences drowsiness, the system sends SMS and emails. The IoT component ensures constant connectivity and data exchange, improving system reliability. Experimental results demonstrate GAN-DD-YOLO's superiority, achieving 98.6% accuracy and faster processing times than other models. By combining GAN and YOLOv8, our approach addresses the need for safe planning in AVs by predicting future states based on current observations, enhancing awareness of imminent hazards. This integration offers a robust solution for the safe path-planning and motion control of AVs, ensuring effective real-time detection and intervention for driver fatigue.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Early drowsiness detection model in autonomous vehicles using GAN and YOLO integration

  • 期刊论文

  • 发表日期：

    2025年10月8日

查看全文
  • Mohammed M. Abo-Zahhad,
  • Sally Elghamrawy,
  • Abdullah Ahmed Hefny,
  • Yasser H. Elawady

摘要

Guaranteeing safe guiding through dynamic environments is a critical mission for autonomous vehicles (AVs). Driver drowsiness is a severe problem that endangers drivers and road safety. Potential accidents can occur because drivers frequently fail to identify indicators of fatigue. This paper proposes a generative adversarial network (GAN) drowsiness detection model using YOLO (GAN-DD-YOLO) for early drowsiness detection. The GAN generates synthetic data that resembles real-world images of drowsy states, such as closed eyes or yawning. The YOLO model was trained on synthetic and real data to increase the model's performance in identifying indications of sleepiness. The proposed model treats each eye separately, which enhances the detection process due to the capability of covering more angles. The proposed model implemented on a Raspberry Pi linked to several sensors and the Internet of things (IoT) network alerts drivers when drowsiness is detected. Whenever the driver experiences drowsiness, the system sends SMS and emails. The IoT component ensures constant connectivity and data exchange, improving system reliability. Experimental results demonstrate GAN-DD-YOLO's superiority, achieving 98.6% accuracy and faster processing times than other models. By combining GAN and YOLOv8, our approach addresses the need for safe planning in AVs by predicting future states based on current observations, enhancing awareness of imminent hazards. This integration offers a robust solution for the safe path-planning and motion control of AVs, ensuring effective real-time detection and intervention for driver fatigue.