Pothole Detection based on Machine Learning and Deep Learning Models

Abstract

Roads are essential for daily transportation worldwide, but their aging and usage patterns can cause deterioration of the road surface, leading to a decline in quality. This deterioration often results in the formation of potholes and cracks on the roads, which can cause damage to vehicles or pose a physical danger to occupants, particularly in underdeveloped countries. Identifying potholes in real-time can help drivers avoid them and prevent accidents. Furthermore, recording their locations and sharing them can assist other drivers and road maintenance organizations take prompt corrective measures. In our attempt to address the issue of pothole detection, The paper aims to combine the latest technological advancements. We aim to develop practical, reliable, adaptable, and modular solutions. To achieve this, we will compare the performance of Random Forest, a machine learning model, with CNN, a deep learning model, in detecting potholes. The experiments were conducted on both models using multiple datasets, and a conclusion was drawn to bring out the benefits of the model with 99% accuracy using CNN and 95% accuracy using Random Forest.