## Cropland-SCD

The pytorch implementation for **MeGNet** in paper "[A Memory Guided Network and A Novel Dataset for Cropland Semantic Change Detection](https://ieeexplore.ieee.org/document/10579791/)" on [IEEE Transactions on Geoscience and Remote Sensing](https://ieeexplore.ieee.org/xpl/RecentIssue.jsp?punumber=36).  

## Requirements
- Python 3.6
- Pytorch 1.7.0


## Datasets
### CropLand Senmantic Change Dection (CropSCD) Dataset
The CropSCD dataset contains 4,141 pairs of high-resolution samples, each with a size of 512×512 and a resolution between 0.5-2 meters.
All images of the samples were randomly collected from rural areas of Guangdong Province, China, while their corresponding labels were precisely annotated through expert visual interpretation.
The dataset contains a total of eight distinct change classes, which include Water, Forest, Plantation, Grassland, Impervious Surface, Greenhouse, Road, and Bare Soil.

- Download the CropSCD Dataset: [OneDrive](https://1drv.ms/u/s!AlRAU4OtVo9zcamTmyagGfylYss?e=LpG2XF) | [Baidu](https://pan.baidu.com/s/1XxNplRS_D4hH9md2NKw7qQ?pwd=b8gg)
- Download the [HRSCD Dataset](https://ieee-dataport.org/open-access/hrscd-high-resolution-semantic-change-detection-dataset)





## Citation

Please cite our paper if you use this code in your work:

```
@ARTICLE{10579791,
  author={Liu, Mengxi and Lin, Simin and Zhong, Yutong and Shi, Qian and Li, Jiaqi},
  journal={IEEE Transactions on Geoscience and Remote Sensing}, 
  title={A Memory Guided Network and A Novel Dataset for Cropland Semantic Change Detection}, 
  year={2024},
  volume={},
  number={},
  pages={1-1},
  keywords={Semantics;Feature extraction;Task analysis;Transformers;Land surface;Soil;Plantations;Remote sensing;semantic change detection;deep learning;memory;Transformer},
  doi={10.1109/TGRS.2024.3421654}}
```