Recommended by

Indexed by

Shadow-light images of simulated 25 classes of surface roughness for automatic classification

Janusz V. Kozubal1, Ahmad Hassanat2*, Ahmad S. Tarawneh3, Roman J. Wróblewski1, Hubert Anysz4*, Jónatas Valença5, Eduardo Júlio5

1 Faculty of Civil Engineering, Wroclaw University of Science and Technology, Wroclaw, Poland

2 Computer Science Department, Mutah University, Karak 61711, Jordan

3 Eötvös Loránd University, Department of Algorithms And Their Applications, Budapest, Hungary

4 Faculty of Civil Engineering, Warsaw University of Technology, Warsaw, Poland

5 CERIS, Instituto Superior Técnico, Universidade de Lisboa, Lisbon, Portugal

* Corresponding authors emails: Ahmad.Hassanat@gmail.com, hubert.anysz@gmail.com
DOI10.24435/materialscloud:y1-jq [version v1]

Publication date: Jun 03, 2022

How to cite this record

Janusz V. Kozubal, Ahmad Hassanat, Ahmad S. Tarawneh, Roman J. Wróblewski, Hubert Anysz, Jónatas Valença, Eduardo Júlio, Shadow-light images of simulated 25 classes of surface roughness for automatic classification, Materials Cloud Archive 2022.68 (2022), doi: 10.24435/materialscloud:y1-jq.


Many relationships important to civil engineering depend on surface roughness (morphology). Examples are the bond strength between concrete layers, the adhesion of a wheel to the pavement, the angle of friction in the soil in contact with a wall surface, and many other cases when we deal with a material with a surface having the characteristics of a Gaussian field. Based on scans of the natural concrete surfaces subjected to different smoothing processes, theoretical models were made. The observed features of the models were grouped into 25 categories belonging to the spherical semivariogram model. Each category is described by two parameters: range (with discrete domain 0.01, 0.04, 0.08, 0.16, 0.32) and upper limits (also with discrete domain 1, 2, 4 , 8, 16) with zero trend. For all combinations of range-limit pairs, homogeneous Gaussian random fields satisfying the spatial dependence of the category were generated in R software by using the RandomFields library. In the final step, the Blender rendering software was used. 1,000 images were taken for each category type using constant angle lighting and moving the camera over the generated surface. Users can download a set of 25,000 images to build or verify the surface classifier. The generated 25 types of surfaces are also provided.

Materials Cloud sections using this data

No Explore or Discover sections associated with this archive record.


File name Size Description
2.9 KiB This is a read me file containing all information about the image dataset
608.3 MiB This is a zip file containing all of the 25000 Shadow-Light Images as described in the Read me file


Files and data are licensed under the terms of the following license: Creative Commons Attribution 4.0 International.
Metadata, except for email addresses, are licensed under the Creative Commons Attribution Share-Alike 4.0 International license.

External references

Journal reference (Paper in which the method is described)
J. Kozubal, A. Hassanat, A. Tarawneh, R. Wróblewski, H. Anysz, J. Valença, E. Júlio, Automatic Strength Assessment of the Virtually Modelled Concrete Interfaces Based on Shadow-Light Images, (in preparation)


concrete Shadow-light images virtual modeling machine learning Classification Deep learning civil engineering

Version history:

2022.68 (version v1) [This version] Jun 03, 2022 DOI10.24435/materialscloud:y1-jq