idomogalla/s1s2_water

S1S2-Water (mirror of v1.0.1)

This repository is an unofficial mirror of the S1S2-Water dataset, created and published by its original authors. I am only re-hosting a copy of version v1.0.1 on Hugging Face to provide a faster and more convenient download alternative to Zenodo, whose transfer speeds can be slow for a dataset of this size (~170 GB).

All credit, ownership, and rights belong to the original authors. Please refer to the original source for the authoritative version, license terms, and any updates.

Original source

• Original dataset (Zenodo, v1.0.1): https://zenodo.org/records/11278238
• Previous version (v1.0.0): https://zenodo.org/records/8314175
• Paper: Wieland, M., Fichtner, F., Martinis, S., Groth, S., Krullikowski, C., Plank, S., Motagh, M. (2023). S1S2-Water: A global dataset for semantic segmentation of water bodies from Sentinel-1 and Sentinel-2 satellite images. IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing. DOI: 10.1109/JSTARS.2023.3333969
• Official code repository: https://github.com/MWieland/s1s2_water

About the dataset

S1S2-Water is a global reference dataset for training, validating, and testing convolutional neural networks for semantic segmentation of surface water bodies in Sentinel-1 and Sentinel-2 satellite images. It consists of 65 triplets of Sentinel-1 and Sentinel-2 images with quality-checked binary water masks. Each sample is complemented with STAC-compliant metadata and a Digital Elevation Model (DEM) raster from the Copernicus DEM.

For full details on the data structure, band descriptions, and preprocessing steps, see the original Zenodo record and the accompanying paper linked above.

Why this mirror exists

Downloading the full dataset directly from Zenodo can be slow depending on your location and network conditions. This mirror simply re-hosts an identical copy of v1.0.1 on Hugging Face's infrastructure to make downloads faster and more reliable. No files have been modified, reprocessed, or altered in any way.

License

This dataset is distributed here exactly as it was published by the original authors. Any use of this data must comply with the license and terms specified in the original Zenodo record: https://zenodo.org/records/11278238

If you are the original author and would like this mirror removed or modified, please contact me directly.

Citation

If you use this dataset, please cite the original paper, not this mirror:

@article{wieland2023s1s2water,
  title={S1S2-Water: A global dataset for semantic segmentation of water bodies from Sentinel-1 and Sentinel-2 satellite images},
  author={Wieland, M. and Fichtner, F. and Martinis, S. and Groth, S. and Krullikowski, C. and Plank, S. and Motagh, M.},
  journal={IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing},
  year={2023},
  doi={10.1109/JSTARS.2023.3333969}
}