Pothole Detection System

Potholes on road surfaces pose significant hazards to vehicles and commuters, contributing to accidents, traffic delays, and vehicle damage. This paper presents a real-time pothole detection system using a lightweight and efficient deep learning model, YOLOv4-tiny, integrated with a live video feed and geolocation logging. A labeled dataset was used to train the model for pothole detection resulting in improving accuracy and relevance in real-world scenarios. The system captures frames from a webcam, identifies potholes with high confidence, highlights them on-screen, saves the corresponding frame, and records geographic coordinates using IP-based geolocation. The results demonstrate reliable performance with real-time detection capabilities, making it a viable solution for smart city infrastructure, road condition monitoring, and preventive maintenance planning. The system is also designed to be computationally efficient, allowing for seamless deployment in resource-constrained environments such as embedded systems or mobile platforms.