DeepCrack: A Deep Hierarchical Feature Learning Architecture for Crack Segmentation

• Resources: [paper] | [code] | [dataset]
• Architecture: based on Holistically-Nested Edge Detection, ICCV 2015, [Paper][code].

• Dataset:

We established a public benchmark dataset with cracks in multiple scales and scenes to evaluate the crack detection systems. All of the crack images in our dataset are manually annotated.

Please note that we own the copyrights to part of original crack images and all annotated maps. Their use is RESTRICTED to non-commercial research and educational purposes.

You can find the dataset in ./dataset, and here are the details:

Folder	Description
train_img	RGB images for training
train_lab	binary annotation for training images
test_img	RGB images for testing
test_lab	binary annotation for testing images

A brief overview on our crack detection dataset:

• Reference:

If you use this dataset for your research, please cite our paper:

@article{liu2019deepcrack,
  title={DeepCrack: A Deep Hierarchical Feature Learning Architecture for Crack Segmentation},
  author={Liu, Yahui and Yao, Jian and Lu, Xiaohu and Xie, Renping and Li, Li},
  journal={Neurocomputing},
  volume={338},
  pages={139--153},
  year={2019},
  doi={10.1016/j.neucom.2019.01.036}
}

If you have any questions, please contact me: yahui.cvrs AT gmail.com without hesitation.