isp-uv-es/seefar-sentinel-taco

SeeFar V0 - Sentinel-2

This is a repackaging, not a new dataset. It is SeeFar V0 - Sentinel-2 by Copernicus / ESA (Sentinel-2), Coastal Carbon (SeeFar assembly; Lowman et al.), 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-SA-3.0-IGO (Copernicus Sentinel data)

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

Citation

Please cite the original work:

@article{lowman2024seefar,
  title   = {SeeFar: Satellite Agnostic Multi-Resolution Dataset for Geospatial Foundation Models},
  author  = {Lowman, James and Zheng, Kelly Liu and Fraser, Roydon and The, Jesse Van Griensven and Valipour, Mojtaba},
  journal = {arXiv preprint arXiv:2406.06776},
  year    = {2024}
}

About the data

The sentinel half of SeeFar V0: 33611 four-band (blue, green, red, NIR) tiles from Sentinel-2 MSI at 10 m (B02, B03, B04, B08). No labels -- this is a multi-resolution self-supervised pretraining corpus.

33,611 samples · splits: test 5,042 · train 23,527 · validation 5,042 · tasks: similarity-search

Packaged as TACO v3.

Full description

Splits. The release's own: 23527 train, 5042 test, 5042 val. The release spells the last one validate; the partition is untouched and split_original keeps its word.

Scope. SeeFar ships four sources in one archive and they are not co-registered: the four id spaces are independent, the four-way filename intersection is 337 of 23,527 in train and 1 in test, and for 120 of those 337 the closest same-name tile in another source is 546 km away, with 0 of 360 pairs within 10 km. Each source is therefore its own collection and nothing may be joined across them; shared_stem records the filename coincidence.

Geometry. Tiles are in per-tile UTM zones, so resolution_m is measured geodesically per sample. The band table is the release's own per-tile wavelength tags, asserted on every file.

Radiometry. Reflectance x 10,000, the mission's quantification value, which is the same constant for L1C and L2A. The processing level is not stated by the release and is declared unknown rather than guessed: the 40 most-vegetated tiles average 0.135 in the blue against 0.024 for this release's Landsat surface-reflectance half, which is the sort of gap Rayleigh scattering makes and leans top-of-atmosphere, but the two samples are different ground, so that is an observation and not a determination.

Tile shapes. 6.6% of the tiles are not 384x384. Over all 2,616 non-standard members, 1,100 are 384x228, 1,082 are 228x384 and 39 are 228x228; they carry real pixels, not one is blank, and they are stored at their native shape rather than padded or dropped.

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:

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

path = snapshot_download("isp-uv-es/seefar-sentinel-taco", repo_type="dataset")
ds = Dataset(path)                  # the repository is the container: parts + .tacocat
plot_sample(ds[0])

or from a local copy:

ds = Dataset("seefar-sentinel")
sample = ds[0]                      # {slot name: SlotValue}, arrays decoded
sample["image"].array.shape

Metadata without decoding anything:

import taco
taco.read("seefar-sentinel/.tacocat")              # one Arrow table, levels joined

Samples

What a sample contains

role	slot	holds	modality	detail
input	image	raster	optical	4 band(s), unit 1, scaled

Licence

CC-BY-SA-3.0-IGO (Copernicus Sentinel data)

Terms of use:

• Adapted material must be shared under CC-BY-SA-3.0-IGO.

Required credits:

• Contains modified Copernicus Sentinel data
• SeeFar: Coastal Carbon (Lowman et al.)

Providers: Copernicus / ESA (Sentinel-2), Coastal Carbon (SeeFar assembly; Lowman et al.)

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.