AIを活用した学校体育館の被災度判定

抄録

Faster R-CNN is used to detect vertical braces, column bases and horizontal roof braces, and the objects are correctly detected in many cases. Accuracy for the damage classification of vertical braces using deep learning is about 72%, accuracy for column bases is about 81%, and accuracy for horizontal roof braces is about 85%. An application software is developed for assessing post-earthquake damage by taking photographs of gymnasium.