Abstract:
Potholes on roads are one of the reasons for road accidents and deploying detection units on all roads for better maintenance is difficult. Deep learning-based pothole detection is a more effective method to easily detect potholes on the road. For that
purpose, different models have been used to detect the potholes with good accuracy and speed. YOLOv5 and YOLOv7 models are trained and then tested on a suitable dataset. We focused on improving the accuracy of the model along with speed using fine-tuning. This will help detect potholes in the city with less time and cost. The main mission of our Final Year Project is to provide easy road inspection methods to road maintaining authorities. This will lessen the damage to vehicles and avoid fatal accidents. This project will also help to provide maintenance cost reduction.
