HSANET：https://chengxihan.github.io/

“HSANET: A HYBRID SELF-CROSS ATTENTION NETWORK FOR REMOTE SENSING CHANGE DETECTION, IGARSS 2025, Chengxi Han, Xiaoyu Su, Zhiqiang Wei, Meiqi Hu, Yichu Xu*, 😋😋😋

[21st Apr. 2023] Release the first version of the HSANet

Requirement

-Pytorch 1.8.0  
-torchvision 0.9.0  
-python 3.8  
-opencv-python  4.5.3.56  
-tensorboardx 2.4  
-Cuda 11.3.1  
-Cudnn 11.3  

Training, Test and Visualization Process

CUDA_VISIBLE_DEVICES=0 python train.py --epoch 50 --batchsize 8  --data_name 'WHU' --model_name 'HSANet'
CUDA_VISIBLE_DEVICES=0 python train.py --epoch 50 --batchsize 8  --data_name 'LEVIR' --model_name 'HSANet'
CUDA_VISIBLE_DEVICES=0 python train.py --epoch 50 --batchsize 8  --data_name 'SYSU' --model_name 'HSANet'
CUDA_VISIBLE_DEVICES=0 python train.py --epoch 50 --batchsize 8  --data_name 'S2Looking' --model_name 'HSANet'

CUDA_VISIBLE_DEVICES=0 python test.py  --data_name 'WHU' --model_name 'HSANet'
CUDA_VISIBLE_DEVICES=0 python test.py  --data_name 'LEVIR' --model_name 'HSANet'
CUDA_VISIBLE_DEVICES=0 python test.py  --data_name 'SYSU' --model_name 'HSANet'
CUDA_VISIBLE_DEVICES=0 python test.py  --data_name 'S2Looking' --model_name 'HSANet'

You can change data_name for different datasets like "LEVIR", "WHU", "SYSU", "S2Looking", "CDD", and "DSIFN".

Test our trained model result

You can directly test our model by our provided HSANet weights in output/WHU, LEVIR, SYSU, S2Looking. Download in Baidu Disk,pwd:2025 😋😋😋

Dataset Download

LEVIR-CD：https://justchenhao.github.io/LEVIR/ , our paper split in Baidu Disk,pwd:2023

WHU-CD：http://gpcv.whu.edu.cn/data/building_dataset.html ,our paper split in Baidu Disk,pwd:2025

SYSU-CD: Our split in Baidu Disk,pwd:2023

S2Looking-CD: Our split in Baidu Disk,pwd:2023

CDD-CD: Our split in Baidu Disk,pwd:2023

DSIFN-CD: Our split in Baidu Disk,pwd:2023

Note: We crop all datasets to a slice of 256×256 before training with it.

Dataset Path Setting

 LEVIR-CD or WHU-CD 
     |—train  
          |   |—A  
          |   |—B  
          |   |—label  
     |—val  
          |   |—A  
          |   |—B  
          |   |—label  
     |—test  
          |   |—A  
          |   |—B  
          |   |—label

Where A contains images of the first temporal image, B contains images of the second temporal image, and label contains ground truth maps.

Acknowledgments

Thanks to all my co-authors Yichu Xu,Meiqi HuThanks for their great work!!

Citation

If you use this code for your research, please cite our papers.

@INPROCEEDINGS{HSANet,
  author={Han, Chengxi and Su, Xiaoyu and Wei, Zhiqiang and Hu, Meiqi and Xu, Yichu},
  booktitle={IGARSS 2025 - 2025 IEEE International Geoscience and Remote Sensing Symposium}, 
  title={HSANET: A Hybrid Self-Cross Attention Network For Remote Sensing Change Detection}, 
  year={2025},
  volume={},
  number={},
  pages={},
  }

Reference

[1] C. HAN, C. WU, H. GUO, M. HU, J.Li, AND H. CHEN, “Change Guiding Network: Incorporating Change Prior to Guide Change Detection in Remote Sensing Imagery,” IEEE J. SEL. TOP. APPL.EARTH OBS. REMOTE SENS., PP. 1–17, 2023, DOI:10.1109/JSTARS.2023.3310208 .

[2] C. HAN, C. WU, H. GUO, M. HU, AND H. CHEN, “HANet: A hierarchical attention network for change detection with bi-temporal very-high-resolution remote sensing images,” IEEE J. SEL. TOP. APPL.EARTH OBS. REMOTE SENS., PP. 1–17, 2023, DOI: 10.1109/JSTARS.2023.3264802.

[3] HCGMNET: A Hierarchical Change Guiding Map Network For Change Detection.