Crack Detection Method Using Coarse Dataset

Abstract

Crack detection on the surface of tunnel lining is an important task to check the condition of the tunnels, one of the representative concrete structure. To achieve a successful result, many deep learning-based methods have been proposed recently. However, the methods require the pure ground truth and it consumes huge resource. To solve the issue, we propose a method to detect crack using both coarse and pure dataset. Experientially, results show that the performance of our method was improved over the previous methods.