2019 Volume 75 Issue 2 Pages I_27-I_35
A bridge inspection needs much cost. It causes a lack of engineer and budget. So some local government couldn’t complete bridge aggressive preventive maintenance in Japan. We focus on image based bridge inspection technique which is research of automated bridge inspection. Some previous researches proposed deep learning based method for detect crack on concrete surface image of bridge and show high accuracy of crack detection. However, the evaluation of robustness of such method to the real bridge surface images is not enough yet. One of the reason of this problem is lack of training data. To make huge dataset needs much cost. In this paper, we proposed a data augmentation method with GAN for solve this problem. We evaluated our method compared with traditional data augmentation method and the results show that our proposed method has high performance than traditional one.
