Awesome

Paddy Rice Maps South Korea (2017~2021)

Binary paddy rice classification from 2017 to 2021 generated by recurrent U-net deep-leaning model. As a result of prediction, it does NOT represent ground truth information, but can be used as a pseudo labeling. (Download)

There are available:

• a raster file with 10 m x 10 m resolution projected by GRS 80 - Korea Central Belt 2010 (EPSG: 5186).

Covering the entire South Korea except Ulleung-gun at the East Sea where no paddy rice exists

Model Used & Validataion

• Recurrent U-net trained with time series Sentinel-1 images and farm map labeling dataset produced by MAFRA (http://data.nsdi.go.kr/dataset/20210707ds00001).
• The trained Sentinel-1 images were composited in the following periods.

• The model was trained with 7,762 patches and validated in 5,180 patches for each patch consists of 256 x 256 pixels.

The above learning material can be downloaded in h5 format (Dataset) (Model) <br/> The dataset is separated into training/valdation data, image/labeling, and part number which can be accessed by key: {tr/va}_{im/lb}_{0~4}<br/> <Python example><br/> <img src="/images/python_example.PNG" width=30%>

• The validation accuracy and Cohen's kappa value are 96.50%, 0.7857 each which were calculated from the 40% of the farm map.

Reference

If you use this dataset, please cite the DOI below 10.5281/zenodo.5845896 (https://doi.org/10.5281/zenodo.5845896)

Acknowledgements

This work was supported by International Research and Development Program of the National Research Foundation of Korea (NRF) funded by the Ministry of Science and ICT under Grant (2021K1A3A1A78097879) and supported by the European Commission under Contract H2020- CALLISTO¹ ( 101004152) by Korea University, South Korea.

Researchers: Hyun-Woo Jo (endeavor4a1@gmail.com), Woo-Kyun Lee (leewk@korea.ac.kr)