This is the code for:

• Pixel-level pavement crack segmentation with encoder-decoder network

This code is tested on:

Ubuntu 20.04

CUDA 11.0 (cuDNN 8.0.5)

Anaconda

To install:

git clone https://github.com/Youzhi1234/EDNet.git
cd EDNet
conda env create -f environment.yml

This code only provide a few images to test. You can replace it with your own datasets.

Open jupyter notebook in brower:

conda activate EDNet
jupyter notebook

Please empty your GPU memory before each run.

First, use '1_decoder.ipynb' to train the decoder network.

Then, use '2_encoder.ipynb' to train the encoder network.

Finally, use '3_EDNet.ipynb' to assemble EDNet, and test it.

We will appreciate your citation

@article{tang2021pixel,
  title={Pixel-level Pavement Crack Segmentation with Encoder-decoder Network},
  author={Tang, Youzhi and Zhang, Allen A and Luo, Lei and Wang, Guolong and Yang, Enhui},
  journal={Measurement},
  pages={109914},
  year={2021},
  publisher={Elsevier}
}