Automated crack assessment: an application on ultra-high-performance concrete (UHPC) panels

This dataset includes 6 images of large-sized concrete cracks, with their respective processed results. Unlike other data used for related studies, which often contain small-sized and well-refined cracks, cracks in this dataset has various features (diagonal orientations, thick cracks with multiple thin branches, etc.). Hence, it is expected that utilizing the submitted dataset will offer a chance for other researchers to test their algorithms and methodologies on data resembling real life data (such as those images collected from buildings or bridges) The dataset containing 6 images was first manually preprocessed for this study [“01.1 Original Images”]. Note that the original images are included in the same folder [“Original Sources” that is under “01.1 Original Images”] along with those preprocessed ones. Next, manually-annotated, pixel-wise ground truths per each image are saved in the next folder [“01.2 Ground Truths”]. The images contained in these two folders will enable other users to assess their algorithms and methods in a more realistic way. In the “02 Pre-processing Results” folder, there are two preliminary results from “02.1 Image Morphological Operation Results” and “02.2 DeepCrack Results.” Finally, the “03 Detection Results” and “04 Skeletonization Results” folders contain the results obtained during this project.