Tackling Road Hazards: Classification and Detection of Crack and Potholes Using Deep Learning

Road Surfaces are the major transportation infrastructure, facilitating economic activities and enabling the smooth mobility of people, goods, and services. Potholes and cracks are significant road hazards that pose risks to road users and vehicles, causing discomfort, vehicle damage, and even leads to accidents. To overcome these issues, a model for classifying road hazards will be developed. Additionally, detection is performed for potholes and cracks. The datasets on which we will be working contain images of normal road surfaces, potholes, and cracks. This model comprises of two key phases. Initially, the ResNet50 algorithm will be used for the classification of road conditions, distinguishing between plain roads, potholes, and cracks. Following this, a YOLOv8 detection algorithm will be implemented to find specific locations of potholes and cracks on the road surface. The existing work for multi classification using ResNet50 got 90% accuracy, which has to be improved by the developed model.