A study on damage evaluation of reinforced concrete column based on deep learning

Abstract

It takes time to carry out the post-earthquake quick inspection of many damaged buildings after large earthquakes. As image classification method of damage level after earthquake, deep learning with convolutional neural network is applied to reinforced concrete columns. We achieve a classification accuracy of 77% by AlexNet and 80% by VGG16 in case of transfer learning. To show the reason of classification, Grad-cam is utilized.