第五部門：応用数理問題（B)

[S05B-06] Basic research on the estimation of building damage using aerial photographs after flood disasters based on deep learning

*Mitsumasa Wada¹, Shiori Kubo¹, Hidenori Yoshida² (1. Kagawa University, 2. Faculty of Engineering and Design, Kagawa University)

Keywords:Deep learning, Flood disaster

Disaster prevention is important as a countermeasure against natural disasters before damages. The reconstruction measures after disasters are also necessary. While the importance of pre-disaster measures is becoming more and more well known, there are many problems with post-disaster recovery measures. In this study, estimation of building damage using aerial photographs after flood disasters based on deep learning is examined. In deep learning, convolutional neural networks and trained data were used. As a result of judging from the aerial photograph after the disaster with the model after learning, the damage situation was grasped by the heat map.