The pytorch implementation for Global-Aware Siamese Network for Change Detection on Remote Sensing Images on ISPRS JOURNAL OF PHOTOGRAMMETRY AND REMOTE SENSING.
The GAS-Net is designed to generate global-aware features for efficient change detection by incorporating the relationships between scenes and foregrounds.
- Python 3.6
- Pytorch 1.2.0
- Download the Levir-CD Dataset
- Download the Lebediv-CD Dataset
The data folder is structured as follows:
├── data/
│ ├── levir_CD/ # Levir-CD dataset
| | ├── train/ # traning set
| | | ├── t1/ #images of time t1
| | | ├── t2/ #images of time t2
| | | ├── label/ #ground truth
| | ├── val/ # validation set
| | | ├── t1/
| | | ├── t2/
| | | ├── label/
| | ├── test/ # testing set
| | | ├── t1/
| | | ├── t2/
| | | ├── label/ #ground truth for evaluation
| | ├── results/ # path to save the model
│ ├── SVCD/
| | ├── leveb/ # Lebediv-CD dataset, have the same structure of the Levir-CD dataset
...
@article{Global2023zhang,
title = {Global-aware siamese network for change detection on remote sensing images},
journal = {ISPRS Journal of Photogrammetry and Remote Sensing},
volume = {199},
pages = {61-72},
year = {2023},
doi = {https://doi.org/10.1016/j.isprsjprs.2023.04.001},
author = {Ruiqian Zhang and Hanchao Zhang and Xiaogang Ning and Xiao Huang and Jiaming Wang and Wei Cui},
}
This code is heavily borrowed from SRCDNet.