AMTNet

The pytorch implementation for AMTNet in paper "An attention-based multiscale transformer network for remote sensing image change detection"on "ISPRS Journal of Photogrammetry and Remote Sensing".

Requirements

• Python 3.9
• Pytorch 1.12

DataSet

• Download the CLCD Dataset
• Download the HRSCD Dataset
• Download the WHU-CD Dataset
• Download the LEVIR-CD Dataset

Prepare datasets into following structure and set their path in train_options.py
├─Train
│  ├─time1
│  │─time2
│  │─label
├─Test
│  ├─time1
│  │─time2
│  │─label

Train

python train.py

All the hyperparameters can be adjusted in train_options.py

model zool

The models with the scores can be downloaded fromBaidu Cloud.

Acknowledgments

This code is heavily borrowed from MSCANet and changer.

Citation

If you find this repo useful for your research, please consider citing the paper as follows:

@article{liu2023attention,
  title={An attention-based multiscale transformer network for remote sensing image change detection},
  author={Liu, Wei and Lin, Yiyuan and Liu, Weijia and Yu, Yongtao and Li, Jonathan},
  journal={ISPRS Journal of Photogrammetry and Remote Sensing},
  volume={202},
  pages={599--609},
  year={2023},
  publisher={Elsevier}
}