A generation of damage classifier for rc partial wall using damage photograph by deep learning

The purpose of this research is to develop a methodology to classify the degree of earthquake damage with no specialists, in order to support the early restoration of the damaged condominium. In order to realize this, we performed fine tuning of the pre-trained convolutional neural network (VGG16), and developed a methodology to identify the damage index from damage photographs of RC partial walls. As a result, some classifiers that could classify the damage index into three ranks (less equals to III, IV, V) with accuracy rates of 91% for the input damage photographs were generated.