Deep Learning-Based Detection of Defects from Images

Pp: 176-182 (7)

* (Excluding Mailing and Handling)

Abstract

Crack detection has vital importance for monitoring and inspection of buildings. It has great significance for structural safety. This is a challenging task for computer vision and machine learning, as cracks only have low-level features for detection. Convolutional Neural Networks (CNN) is a very promising framework for crack detection from images with high accuracy and precision. This paper is based on a deep-learning methodology to detect and recognize structural defects. The dataset is split into training and testing data which is used to train the model. Then this trained model is used to recognize and classify cracks in images. The dataset consists of concrete crack images. The data set used has two categories, images with cracks and without cracks. A Convolutional Neural Network model using Pytorch will be fit to predict the images if the images have any cracks or not. This paper compares the accuracy of various models.

Keywords: CNN, Defect detection, Image processing, Machine learning.

Cite as