Real Time Implementation of Pothole Detection Using Deep Learning

One of the emerging problem roads faces is worsened road conditions and potholes. Many physical conditions like rains, road calamities or wear and tear make the road problematic to drive. Also, while driving in the darkness just the headlights might not be enough support for driver. Unpredicted potholes on road may cause more accidents. Also due to worse road environments, fuel intake of the vehicle exceeds and causing loss of expensive fuel. All these explanations craving that there is necessity of making a model which can predict the potholes in advance and convey it to driver. Therefore, in this work, we address this issue with the help of deep learning model and Raspberry Pi based hardware setup. Yolov8 has been employed to detect all types of potholes whereas Raspberry Pi is used to implement it in real time. The experimentation is carried out on a potholes image dataset with multiple variations. The testing has been done on real-time video captured through a camera setup connected to Raspberry Pi. The experimentation results proved the strength of the proposed approach for pothole detection with 92% detection accuracy at 26.2FPS.