U-Net-Based CNN Architecture for Road Crack Segmentation

Many studies on the semantic segmentation of cracks using machine learning (ML) technique can be found in relevant literature. To date, results obtained are quite good, but often accuracy trained model and evaluated traditional metrics only, most cases, goal is to detect only occurrence cracks. Particular attention should paid thickness segmented crack since, road pavement maintenance, width main parameter one that characterizes severity levels. The aim our study optimize process through implementation a modified U-Net model-based algorithm. For this, Crack500 dataset used, then compared with those from algorithm, which currently accurate performant promising accurate, as findings shape very close reality.