SmartDriveGuard: AI-Powered Real-Time Pothole Detection in Autonomous Vehicles Using YOLO-Based Object Detection

  • Unique Paper ID: 166839
  • Volume: 11
  • Issue: 2
  • PageNo: 2241-2248
  • Abstract:
  • Pothole detection is a critical component in maintaining road safety and minimizing vehicle damage. Traditional manual inspection methods are labor-intensive, time-consuming, and often subject to human error. With the advancements in computer vision and deep learning, automated detection systems offer a promising solution. This paper explores the application of various YOLO (You Only Look Once) models for real-time pothole detection. We evaluate the performance of different YOLO architectures, including YOLOv3, YOLOv4, and YOLOv5, on a dataset consisting of road images with and without potholes. The models are trained and tested to determine their accuracy, speed, and efficiency in identifying potholes under various lighting and weather conditions. The real-time detection capability is assessed using metrics such as frames per second (FPS) and average precision (AP). Our results indicate that YOLOv5 outperforms its predecessors in both accuracy and speed, making it the most suitable candidate for real-time applications. YOLOv4, while slightly less accurate, offers a good balance between detection speed and precision. YOLOv3, despite being the oldest model in the comparison, still provides reliable detection but falls short in real-time performance. The study highlights the potential of these models to be integrated into mobile or vehicular systems for continuous monitoring of road conditions. We discuss the implications of deploying these systems in smart city infrastructure, emphasizing the benefits of proactive maintenance and enhanced road safety. Future work will focus on optimizing these models for deployment on edge devices and improving their robustness against various environmental challenges.

Related Articles