Masonry Crack Detection Dataset

The dataset contained in this repository is a collection of crack images and their annotated bounding-box annotations. Currently, the dataset only accounts for Crack defects, although it could be expanded in the future. The dataset was targeted to detection tasks, hence the bounding-box labeling.

[Paper] [BibTeX]

Disclaimer

It is possible some defects are not optimally labelled as I do not have civil engineering qualifications. If you find any errors, please open an Issue with the details. Thank you!

Num Images	Crack Instances
3291	5954

Images

Crack images of size 224 x 224 pixels.

Annotations

The bounding-box labels are structured following the format used for YOLO format. Each image in the dataset has a corresponding .txt file containing the objects in the image. The text files are formatted as:

<object-id> <x-center> <y-center> <width> <height>

• object-id: integer representing the class of the object. This should start from 0 and increase by 1 for each new object class. Currently, this dataset only contains 'Crack' images, therefore <object-id> is always 0. In the event of adding more defects, this will be updated accordingly.

• x/y-center: coordinates of the bounding-box centre, normalised by the width and height of the image. Values should range within [0,1].

• width/height: dimensions of the bounding-boxes, normalised by the width and height of the image. Values should range within [0,1].

Examples

Contributing

There are several tasks that one could do to improve the current dataset such as, segmentation, add more defects, error spotting... If you would like to contribute to the project, please follow the guidelines:

1. Fork the repo and create your branch from 'main'.
2. If you've added more images, make sure that their size is consistent with the rest of the dataset (224 x 224).
3. If you've added more bounding-box annotations, ensure they follow YOLO formatting [YOLO].
4. For segmentation, consider using Roboflow labeling tool.

Issues

Please ensure your description is clear and has sufficient instructions to be able to reproduce the issue.

Citing this dataset

If you find this dataset useful, please consider giving a star ⭐ and citation 🦖:

@inproceedings{marin21,
  author = {Marin, B. and Brown, K. and Erden, M. S.},
  booktitle = {2021 IEEE 17th International Conference on Automation Science and Engineering (CASE)}, 
  title = {Automated Masonry crack detection with Faster R-CNN}, 
  year = {2021},
  volume = {},
  number = {},
  pages = {333-340},
  DOI = {10.1109/CASE49439.2021.9551683}
  }