README.md

---
license: gpl-3.0
task_categories:
- image-segmentation
tags:
- synthetic
pretty_name: Synthetic masonry surfaces for crack segmentation
size_categories:
- 1K<n<10K
---

# Synthetic masonry surfaces

A synthetic masonry crack dataset. This dataset is meant to be used with image segmentation models aimed at segmenting cracks in masonry surfaces, for instance using [this](https://github.com/dimitrisdais/crack_detection_CNN_masonry) network.
This dataset was generated using a [3D dataset generation framework for masonry surfaces](https://github.com/DavidHidde/cracked-surface-generation) and is available under the GPL-3.0 license.

## Dataset structure

This dataset consist of 2 subdatasets: a dataset generated using a 'patches' rendering approach and one using a regular rendering approach. The patches rendering approach renders images at a higer resolution and then cuts the result into smaller patches. Both dataset archives consist of both 4058 input images and generated ground-truth masks.

## Citation

We kindly ask you to cite this dataset using the assigned DOI if you use it in your work: https://doi.org/10.57967/hf/2214