Vision-Based Deep Learning Algorithm for Detecting Potholes

Abstract Potholes on roads pose a major threat to motorists. Driving over pothole has the potential cause serious damage vehicle, which in turn may result fatal accidents. Currently, many detection methods exist. However, these do not utilize deep learning techniques detect real-time, determine location thereof and display its map. The success of determining an effective method, includes aforementioned techniques, is dependent acquiring large amount data, including images potholes. Once adequate data had been gathered, were processed annotated. next step was algorithms could be utilized. Three different models, Faster R-CNN, SSD YOLOv3 trained custom dataset containing potholes network produces best results for real-time detection. It revealed that produced most accurate performed with average time only 0.836 s per image. final system, integrated cloud maps service, can created allow drivers avoid