isp-uv-es/l8biome-taco

<p align="center"><img src="assets/logos.png" width="546" alt="Image and Signal Processing Group, Universitat de València · Elliot"></p>

L8 Biome cloud validation

This is a repackaging, not a new dataset. It is L8 Biome cloud validation by U.S. Geological Survey, torchgeo (torchgeo/l8biome), converted to TACO with its data unchanged. 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{usgs2016l8biome,
  author    = {U.S. Geological Survey},
  title     = {L8 Biome Cloud Validation Masks},
  year      = {2016}, publisher = {USGS ScienceBase},
  doi       = {10.5066/F7251GDH}
}
@article{foga2017cloud,
  title   = {Cloud detection algorithm comparison and validation for operational Landsat data products},
  author  = {Foga, Steve and Scaramuzza, Pat L. and Guo, Song and Zhu, Zhe and Dilley, Ronald D. and Beckmann, Tim and Schmidt, Gail L. and Dwyer, John L. and Hughes, M. Joseph and Laue, Brady},
  journal = {Remote Sensing of Environment}, volume = {194},
  pages   = {379--390}, year = {2017},
  doi     = {10.1016/j.rse.2017.03.026}
}

About the data

The USGS Landsat 8 cloud-cover-assessment validation set: 96 full OLI/TIRS scenes, twelve in each of 8 biomes (barren, forest, grasscrops, shrubland, snowice, urban, water, wetlands), each with a hand-drawn four-class cloud mask covering the whole scene.

96 samples · splits: test 10 · train 62 · validation 24 · tasks: semantic-segmentation

Packaged as TACO v3.

<details> <summary>Full description</summary>

Bands. Eleven channels, B1-B11 on the 30 m grid, including the resampled 15 m panchromatic band.

Radiometry. The pixels in this mirror are uint8 while their own MTLs declare QUANTIZECALMAX = 65535. The release contradicts itself and the pixels win: the slot is requantised with no approximate scale, and the MTL's radiance and reflectance coefficients are carried under names that say they apply to the 16-bit original. The requantisation is not a per-scene contrast stretch -- dn_max shows scenes that never approach 255 -- but the exact mapping needs the *_raw.tar.gz siblings, which are not held.

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.

Splits. The release publishes none and none is invented; biome and the WRS path/row are what a biome-disjoint split would be built on.

Resolution. Measured geodesically per scene.

</details>

Getting started

git clone https://github.com/OscarPellicer/taco
pip install -e "taco[ml]"

Read it straight from the Hub:

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

path = hf_hub_download("isp-uv-es/l8biome-taco", "l8biome.tacozip", repo_type="dataset")
ds = Dataset(path)
plot_sample(ds[0])

or from a local copy:

ds = Dataset("l8biome.tacozip")
sample = ds[0]                      # {slot name: SlotValue}, arrays decoded
sample["image"].array.shape

Metadata without decoding anything:

import taco
taco.read("l8biome.tacozip")              # one Arrow table, levels joined

Samples

[image] [image] [image] [image] [image]

What a sample contains

Licence

CC0-1.0

Providers: U.S. Geological Survey, torchgeo (torchgeo/l8biome)

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.