Pothole Detection using Roboflow Convolutional Neural Networks
D. Deepa, A. Sivasangari, Rahul Roonwal, Rajeev Nayan
Abstract
India has a vast network of roads which connects the various cities and villages, due to which its maintenance becomes a challenging task and also many accidents takes place because of the damaged roads and also the accidents are increasing every year because of the increase in the number of potholes. As the inspection of these roads for damages is done manually, a significant amount of time is consumed, cost increases as it is a labor intensive process and also the possibility of human error increases. Also, as the road network is very large, it is not feasible for the Authority travel to all the places for inspection. In order to solve this problem and to increase the efficiency we can use various Image Processing techniques (such as CNN, YOLOv5) in order to find the damages and then comparing the same to find the technique with maximum accuracy. This data can further be sent to the government for repairing the roads and this helps in reducing accidents and further ensuring the safety of the citizens especially during rainy season. By using these techniques, the problem of damaged roads can be solved and also a lot of resources can be saved by repairing the roads in the right time and also by reducing the labor cost.