A global dataset for cloud and cloud shadow semantic understanding

• Introduction  • Instructions  • Citation  • Credits

Introduction

In order to cover EO benchmarking requirements, we join to each IP the results of eight of the most popular CD algorithms:

• Fmask4: Function of Mask. We use the MATLAB implementation code via Linux Docker containers. This resource is available in https://github.com/GERSL/Fmask.

• Sen2Cor: The Scene Classification (SC), which provides a semantic pixel-level classification map. The SC maps are obtained from the “COPERNICUS/S2_SR” GEE dataset.

• sen2cloudless: Single-scene CD algorithm created by Sentinel-Hub using LightGBM decision tree model\cite{Ke2017}. This cloud mask is available in the “COPERNICUS/S2_CLOUD_PROBABILITY” GEE dataset.

• López-Puigdollers et al. 2021: UNet with two different SEN2 band combination RGBI (B2, B3, B4, and B8) and RGBISWIR (B2, B3, B4, B8, B11, and B12) trained on Biome-8. This resource is available in https://github.com/IPL-UV/DL-L8S2-UV.

• qa60: Cloud mask embeds in SEN2 Level-1C products. The cloud mask are obtained from the “COPERNICUS/S2” GEE dataset.

• kappaMask: UNet with two distinct settings: all Sentinel-2 L1C bands and all Sentinel-2 L2A bands except the Red Edge 3 band. It was trained in an extension of the Sentinel-2 Cloud Mask Catalogue.

Instructions

1. Go to the model folder.
2. Follow the instructions.

Citation

@article{aybar2022cloudsen12,
  title={CloudSEN12-a global dataset for semantic understanding of cloud and cloud shadow in Sentinel-2},
  author={Aybar, Cesar and Ysuhuaylas, Luis and Loja, Jhomira and Gonzales, Karen and Herrera, Fernando and Yali, Roy and Flores, Angie and Diaz, Lissette and Cuenca, Nicole and Espinoza, Wendy and Prudencio, Fernando and Llactayo, Valeria and Montero, David and Sudmanns, Martin and Tiede, Dirk and Mateo-García, Gonzalo and Gómez-Chova, Luis},
  year={2022},
  publisher={EarthArXiv}
}

Acknowledgment

This project gratefully acknowledges:

for computing resources