CRACK DETECTION USING DEEP LEARNING WITH ATTENTION MECHANISMS

Abstract

The aging of concrete structures such as bridges and tunnels has become a problem. In order to reduce the labor required for inspection, automatic detection methods have been developed, especially for cracks. However, for the semantic segmentation of cracks, a large number of supervised images specifying cracks at the pixel level are required, which is very difficult to create. In this study, we developed a new method for the semantic segmentation of cracks while reducing the cost of creating training dataset. Then, the accuracy of the method was evaluated using concrete images, and it was confirmed that the accuracy of the proposed method was similar to the existing detection methods.