Deep Learning Approaches for Crack Detection in Bridge Concrete Structures

Convolutional Neural Networks are among the most effective algorithms for image analysis applications. However, the accuracy of the algorithms depends on the availability of powerful computational resources and the quality of the images used to train the models. This paper investigates ways to build robust models to detect cracks in concrete structures using low resolution images and third-party datasets. Our experiments show that reducing image sizes by a factor of 4 does not significantly impact the accuracy. This is helpful to shorten execution time and hence lower cloud service costs. It is also observed that applying a model trained on one image dataset to detect cracks in images from a different source is not a trivial task.