Intelligent Deep Learning based Pothole Detection and Reporting System
R Rohitaa, Sangoju Shreya, R. Amutha
Abstract
Road maintenance is crucial to prevent accidents due to potholes. Manual assessment of the road condition is a difficult task as it is a tedious process and requires lot of manpower. Thus, there is an increasing requirement for an automatic pothole identification system. In this paper, a solution is proposed to detect the potholes on roads automatically using the deep learning algorithms. Three deep learning namely, Convolutional Neural Network (CNN), Mask Region-based Convolutional Neural Network (Mask RCNN) and You Only Look Once (YOLOv3) are trained and tested with a dataset. The results of the three models are compared using evaluation metrics. This system also assembles hardware components for reporting the potholes in order to take actions for their repair and maintenance and warns the presence of potholes to drivers.