Real Time Pot Hole Detection for Safe Roadways

As roadways are developing at an increasingly faster rate with increase in rate of road users, smart and safe transportation systems are required. In this view, an automated pothole identification system based on YOLOv5 (You Only Look Once) method is presented in this paper. Using live video feed and still photos, the technology is intended to find and locate potholes on roads. The YOLOv5 model is trained and fine-tuned with multiple hyper parameters to achieve high accuracy in pothole identification. The dataset is collected and annotated using Roboflow. To improve road maintenance and safety, the system is owned to track and detect potholes on roads in real-time. Performance of proposed system is measured using evaluation parameters such as mAP, precision, F1 score and recall. Presented approach highlights the significance of utilizing cutting-edge technologies for safer and more effective road maintenance while demonstrating the usefulness and promise of employing YOLOv5 for automated pothole detection.