Automated Pothole Detection using Transfer Learning

Road irregularities like potholes, which are common and persistent, provide a serious problem for both drivers and the government. In order to address this long-standing problem, the research study examines a novel strategy that combines transfer learning with ResNet (Residual Neural Network) methodologies. The main cause of pothole formation is unfavourable weather, which gradually erodes road surfaces and leaves voids behind. Potholes are created as a result of these voids expanding over time and are frequently hidden by flooding or rain during bad weather. Current government responses frequently suffer from delays and resource shortages and rely on manual inspections and reactive repair plans. The paper suggests developing real-time, inexpensive pothole detecting systems with a remarkable accuracy rate of 98.24 percent by utilising ResNet and transfer learning. These solutions seamlessly incorporate the locations of discovered potholes, along with geo-tagging tools that accurately identify each pothole's coordinates and map them to Google Maps for public access and awareness. This proactive strategy enables authorities to quickly identify and prioritise maintenance activities, greatly improving road safety, lowering the cost of vehicle maintenance, and efficiently allocating resources. In the end, while being specifically created for use by government officials, this technology-driven strategy offers a significant improvement over current methods and stands to benefit the public by raising the calibre of transportation infrastructure.