isp-uv-es/l7irish-taco

L7 Irish cloud validation

This is a repackaging, not a new dataset. It is L7 Irish cloud validation by U.S. Geological Survey, 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: CC0-1.0

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

Citation

Please cite the original work:

@misc{usgs2016l7irish,
  author    = {Scaramuzza, Pat and Dwyer, John},
  title     = {L7 Irish Cloud Validation Masks},
  publisher = {U.S. Geological Survey},
  year      = {2016},
  doi       = {10.5066/F7XD0ZWC}
}
@article{scaramuzza2012ldcm,
  title   = {Development of the Landsat Data Continuity Mission Cloud-Cover Assessment Algorithms},
  author  = {Scaramuzza, Pasquale L. and Bouchard, Michelle A. and Dwyer, John L.},
  journal = {IEEE Transactions on Geoscience and Remote Sensing},
  volume  = {50},
  number  = {4},
  pages   = {1140--1154},
  year    = {2012},
  doi     = {10.1109/TGRS.2011.2164087}
}

About the data

The USGS Landsat 7 cloud-cover-assessment validation set: 206 full ETM+ scenes over 9 climate zones (austral, boreal, mid_latitude_north, mid_latitude_south, polar_north, polar_south, subtropical_north, subtropical_south, tropical), each with a hand-drawn four-class cloud mask covering the whole scene.

206 samples · splits: test 21 · train 135 · validation 50 · tasks: semantic-segmentation

Packaged as TACO v3.

Full description

Bands. Nine channels on the 30 m grid: B1-B5, the two gain halves of the thermal band 6, B7 and the resampled panchromatic B8.

Radiometry. image is the release's native 8-bit DN, declared digital_number: the MTLs publish an exact radiance conversion, but per scene and per band, so its coefficients are carried as metadata (radiance_lmin/radiance_lmax against qcal_min/qcal_max). Two layers computed from it sit beside it: toa, TOA reflectance of B1-B5 and B7 (reflectance x 1e4, int16), and bt, band 6 low-gain brightness temperature (kelvin x 100, uint16). Each is per scene from that scene's LMIN/LMAX/QCALMIN/QCALMAX, sun elevation and date, with Chander et al. (2009) ESUN and K1/K2. DN 0 is fill in all three.

Target. The mask's raw codes 0/64/128/192/255 are remapped to 0..4 = Fill / Cloud Shadow / Clear / Thin Cloud / Cloud, with Fill as the ignore index. 0 is also the imagery's fill, being below the product's own QCALMIN of 1.

Splits. This is a validation reference set rather than a training benchmark, and the release publishes no split. split is a constructed 80/10/10 partition that keeps each WRS path/row on one side; it is not zone-disjoint, and zone is what a zone-disjoint split would be built on.

Resolution. Measured geodesically per scene rather than read off the transform.

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/l7irish-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("l7irish")
sample = ds[0]                      # {slot name: SlotValue}, arrays decoded
sample["image"].array.shape

Metadata without decoding anything:

import taco
taco.read("l7irish/.tacocat")              # one Arrow table, levels joined

Samples

What a sample contains

role	slot	holds	modality	detail
input	image	raster	multispectral	9 band(s), digital_number
input	toa	raster	multispectral	6 band(s), unit 1, scaled
input	bt	raster	thermal	1 band(s), unit K, scaled
target	cloud	mask	label_raster	5 classes

Licence

CC0-1.0

Providers: U.S. Geological Survey

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.