orion-ai-lab/Kuro-Siwo-Webdataset

Kuro Siwo webdatasets

Paper | GitHub |

Dataset Details

Dataset Description

Kuro Siwo is a global multi-temporal SAR dataset for rapid flood mapping. It contains 43 flood events in 6 continents and 3 climate zones, over the period 2015-2022. The annotations have been produced through meticulous photointerpretation by a team of experts, at 10m spatial resolution. For each flood event, we provide one Sentinel-1 post-flood and two Sentinel-1 pre-flood captions, along with a Digital Elevation Model (DEM). For research purposes, we opted to include both GRD and SLC products.

This particular repo contains the webdatasets for BlackBench, the benchmark that accompanies Kuro Siwo in the corresponding publication. We provide the train and test sets for both GRD and SLC products of the labelled component of Kuro Siwo.

• Funded by: ThinkingEarth (grant agreement No 101130544) and MeDiTwin (grant agreement No 101159723) of the European Union’s Horizon Europe research and innovation programme
• License: CC BY

Dataset Structure

Sample fields

Each sample in the GRD webdatasets has the following fields:

• flood_vv.npy (float32): The VV band of the post-flood image
• flood_vh.npy (float32): The VH band of the post-flood image
• sec1_vv.npy (float32): The VV band of the first pre-flood image
• sec1_vh.npy (float32): The VH band of the first pre-flood image
• sec2_vv.npy (float32): The VV band of the second pre-flood image
• sec2_vh.npy (float32): The VH band of the second pre-flood image
• dem.npy (float32): The DEM
• info.json: Metadata on this particular sample
• mask.npy (float32): The label (0: no water, 1: permanent water, 2: flood)
• valid_mask.npy (float32): A mask depicting the valid pixels (0: invalid, 1: valid)

Accordingly, each sample in the SLC dataset has the following fields:

• flood.npy (float32): The post-flood image *
• sec1.npy (float32): The first pre-flood image *
• sec2.npy (float32): The second pre-flood image *
• dem.npy (float32): The DEM
• info.json: Metadata on this particular sample
• mask.npy (float32): The label (0: no water, 1: permanent water, 2: flood)
• valid_mask.npy (float32): A mask depicting the valid pixels (0: invalid, 1: valid)

• In the SLC component, each image has 4 channels: Phase VV, Amplitude VH, Amplitude VV, Phase VH.

Data splits

We construct a challenging evaluation framework, selecting 10 flood events across the globe as testing sites, covering a wide range of environmental conditions representing all six continents and three major climate zones featured in Kuro Siwo.

The train and test sets for the GRD component are in the train_GRD and test_GRD folders respectively, whereas the train and test sets for the SLC components are in the train_SLC and test_SLC folders respectively.

Citation

BibTeX:

@inproceedings{NEURIPS2024_43612b06,
 author = {Bountos, Nikolaos Ioannis and Sdraka, Maria and Zavras, Angelos and Karavias, Andreas and Karasante, Ilektra and Herekakis, Themistocles and Thanasou, Angeliki and Michail, Dimitrios and Papoutsis, Ioannis},
 booktitle = {Advances in Neural Information Processing Systems},
 editor = {A. Globerson and L. Mackey and D. Belgrave and A. Fan and U. Paquet and J. Tomczak and C. Zhang},
 pages = {38105--38121},
 publisher = {Curran Associates, Inc.},
 title = {Kuro Siwo: 33 billion m\^{}2 under the water. A global multi-temporal satellite dataset for rapid flood mapping},
 url = {https://proceedings.neurips.cc/paper_files/paper/2024/file/43612b0662cb6a4986edf859fd6ebafe-Paper-Datasets_and_Benchmarks_Track.pdf},
 volume = {37},
 year = {2024}
}

APA:

Bountos, N. I., Sdraka, M., Zavras, A., Karavias, A., Karasante, I., Herekakis, T., Thanasou, A., Michail, D. & Papoutsis, I. (2024). Kuro Siwo: 33 billion $ m^ 2$ under the water. A global multi-temporal satellite dataset for rapid flood mapping. Advances in Neural Information Processing Systems, 37, 38105-38121.

Dataset Card Authors

Maria Sdraka