zyxu1996 / HRCNet-High-Resolution-Context-Extraction-Network

SOTA method on Potsdam and Vaihingen datasets, based on HRNet.

HRCNet: High-Resolution Context Extraction Network for Semantic Segmentation of Remote Sensing Images

This repository is a PyTorch implementation for our Remote Sensing paper.

Usage

Version pytorch >=1.1, we advise torch1.6.
Training

sh auto_train.sh

Testing

python test.py

Restore image

python produce_result.py

Count metrics

python count_metrics.py

Acknowledgments

Thanks the ISPRS for providing the Potsdam and Vaihingen datasets.

Citation

@article{xu2020hrcnet,
  title={HRCNet: high-resolution context extraction network for semantic segmentation of remote sensing images},
  author={Xu, Zhiyong and Zhang, Weicun and Zhang, Tianxiang and Li, Jiangyun},
  journal={Remote Sensing},
  volume={13},
  number={1},
  pages={71},
  year={2020},
  publisher={MDPI}
}

Other Links

• Efficient Transformer for Remote Sensing Image Segmentation
• CCTNet: Coupled CNN and Transformer Network for Crop Segmentation of Remote Sensing Images