Selective Feature Fusion and Irregular-Aware Network for Pavement Crack Detection

👀Introduction

This repository contains the code for our paper Selective Feature Fusion and Irregular-Aware Network for Pavement Crack Detection. [Paper]

💡Requirements

Environment requirements:

• CUDA 11.7

• Python 3.8

Dependency requirements:

• numpy 1.20.0
• torch 1.13.1
• torchaudio 2.0.2
• torchvision 0.14.1

📦Usage

Training

You can modify the training parameters in the train_SFIAN.sh file and run it with the following command:

bash ./scripts/train_SFIAN.sh 

Testing

You can modify the test parameters in the test_SFIAN.sh file and run it with the following command:

bash ./scripts/test_SFIAN.sh 

Evaluation

Calculate ODS, OIS, P, R, F1, mIoU metrics:

cd eval
python evaluate.py

Calculate params, FLOPs metrics:

cd eval
python flops.py

📌BibTeX & Citation

If you find this code useful, please consider citing our work:

@article{cheng2023selective,
  title={Selective feature fusion and irregular-aware network for pavement crack detection},
  author={Cheng, Xu and He, Tian and Shi, Fan and Zhao, Meng and Liu, Xiufeng and Chen, Shengyong},
  journal={IEEE Transactions on Intelligent Transportation Systems},
  year={2023},
  publisher={IEEE}
}

@article{liu2024staircase,
  title={Staircase Cascaded Fusion of Lightweight Local Pattern Recognition and Long-Range Dependencies for Structural Crack Segmentation},
  author={Liu, Hui and Jia, Chen and Shi, Fan and Cheng, Xu and Wang, Mianzhao and Chen, Shengyong},
  journal={arXiv preprint arXiv:2408.12815},
  year={2024}
}