CONSIDERATION OF CRACK WIDTH MEASUREMENT OF REINFORCED CONCRETE STRUCTURES BY USING PLURAL DEEP LEARNING MODELS

Abstract

Recently, after a huge earthquake, reinforced concrete buildings were not available or demolished due to sever damages. Therefore, a damage assessment becomes important; hence, measuring damages from images is one of the most useful techniques. In this study, crack widths of the non-structural wall specimens were measured by using plural deep learning model. By the models which provide the extremely small values of Accuracy and Precision, cracks could not be predicted. While, the deep learning model, in which the values for Recall and F1Score were high, could properly identify the cracks; then, the crack width was reasonably measured.