isp-uv-es/copernicus-bench-cloud-s3-taco

Copernicus-Bench Cloud-S3

This is a repackaging, not a new dataset. It is Copernicus-Bench Cloud-S3 by Technical University of Munich (Copernicus-FM / Copernicus-Bench), ESA Copernicus / Sentinel-3, converted to TACO. Pixel values and labels are kept as released except where the description below says otherwise. All credit belongs to the original authors: if you use it, please cite them and follow their licence.

original dataset · paper · licence: CC-BY-4.0

Repackaged into TACO by the Image and Signal Processing Group (ISP), Universitat de València, within the ELLIOT project.

Citation

Please cite the original work:

@inproceedings{wang2025copernicusfm,
  title     = {Towards a Unified {Copernicus} Foundation Model for Earth Vision},
  author    = {Wang, Yi and Xiong, Zhitong and Liu, Chenying and Stewart, Adam J. and Dujardin, Thomas and Bountos, Nikolaos Ioannis and Zavras, Angelos and Gerken, Franziska and Papoutsis, Ioannis and Leal-Taix{\'e}, Laura and Zhu, Xiao Xiang},
  booktitle = {2025 IEEE/CVF International Conference on Computer Vision (ICCV)},
  pages     = {9888--9899},
  year      = {2025},
  doi       = {10.1109/ICCV51701.2025.00922}
}

About the data

Cloud segmentation on Sentinel-3 OLCI Full-Resolution radiance: 1995 scenes of 21 bands at 256x256. Task L1 of Copernicus-Bench.

1,995 samples · splits: test 399 · train 1,197 · validation 399 · tasks: semantic-segmentation

Packaged as TACO v3.

Full description

Target. Both the official 6-class mask (invalid, clear, cloud-ambiguous, cloud-sure, cloud shadow, snow/ice) and the authors' binary reduction are carried.

• The authors' binary mask groups cloud shadow with clear. The canonical ml:binary reduction declared on the 6-class mask does not, and treats shadow as contaminated.
• NaN in the source marks unobserved pixels and is declared as nodata here; the source files declare none.

Splits. The official splits, preserved.

Getting started

git clone --recursive https://github.com/OscarPellicer/taco
pip install -e "taco/python[ml]"     # builds the reader: C++23, CMake, Ninja, pkg-config, libcurl >= 7.83, OpenSSL >= 3

Read it straight from the Hub:

import os
from huggingface_hub import hf_hub_download, snapshot_download
from taco.ml import Dataset, plot_sample

path = hf_hub_download("isp-uv-es/copernicus-bench-cloud-s3-taco", "copernicus-bench-cloud-s3.zip", repo_type="dataset")
ds = Dataset(path)
plot_sample(ds[0])

or from a local copy:

ds = Dataset("copernicus-bench-cloud-s3.zip")
sample = ds[0]                      # {slot name: SlotValue}, arrays decoded
sample["s3"].array.shape

Metadata without decoding anything:

import taco
taco.read("copernicus-bench-cloud-s3.zip")              # one Arrow table, levels joined

Samples

What a sample contains

role	slot	holds	modality	detail
input	s3	raster	multispectral	21 band(s), unit mW m-2 sr-1 nm-1, physical
target	cloud_multi	mask	label_raster	6 classes
target	cloud_binary	mask	label_raster	3 classes

Licence

CC-BY-4.0

Providers: Technical University of Munich (Copernicus-FM / Copernicus-Bench), ESA Copernicus / Sentinel-3

Acknowledgements

TACO was designed by César Aybar and is specified at https://asterisk.coop/taco/spec/.

Built by Oscar Pellicer within the ELLIOT project at the Image and Signal Processing Group (ISP), Universitat de València.