CrackFormer

CrackFormer: Transformer Network for Fine-Grained Crack Detection, Huajun Liu, Xiangyu Miao, Christoph Mertz, Chengzhong Xu, Hui Kong; ICCV2021

CrackFormer Network for Pavement Crack Segmentation, Huajun Liu, Jing Yang, Xiangyu Miao, Christoph Mertz, Hui Kong; IEEE TITS 2023

usage

datasets

Download the cracktree,[crackls315](Deepcrack: Learning hierarchi- cal convolutional features for crack detection.),stone331,crack537 dataset and the file follows the following structure.

|-- datasets
    |-- crack315
        |-- train
        |   |-- train.txt
        |   |--img
        |   |   |--<crack1.jpg>
        |   |--gt
        |   |   |--<crack1.bmp>
        |-- valid
        |   |-- Valid_image
        |   |-- Lable_image
        |   |-- Valid_result
        ......

train.txt format

./dataset/crack315/img/crack1.jpg ./dataset/crack315/gt/crack1.bmp
./dataset/crack315/img/crack2.jpg ./dataset/crack315/gt/crack2.bmp
.....

Valid

Change the 'pretrain_dir','datasetName' and 'netName' in test.py

python test.py

Pre-trained model

Download the trained model.

|-- model
    |-- <netname>
        |   |-- <trained_model.pkl>
        ......

dataset	Pre-trained Model
cracktree	Link
crackls315	Link
stone331	Link
crack537	Link

                                                   |

Acknowledgments

• We thank the anonymous reviewers for valuable and inspiring comments and suggestions.

Reference

@InProceedings{Liu_2021_ICCV,
author    = {Liu, Huajun and Miao, Xiangyu and Mertz, Christoph and Xu, Chengzhong and Kong, Hui},
title     = {CrackFormer: Transformer Network for Fine-Grained Crack Detection},
booktitle = {Proceedings of the IEEE/CVF International Conference on Computer Vision (ICCV)},
month     = {October},
year      = {2021},
pages     = {3783-3792}
}

@article{Liu2023CrackFormer,
title={CrackFormer Network for Pavement Crack Segmentation},
author={Huajun Liu and Jing Yang and Xiangyu Miao and Christoph Mertz and Hui Kong},
journal={IEEE Transactions on Intelligent Transportation Systems},
volume={24},
number={9},
pages={9240-9252},
year={2023},
publisher={IEEE}
}