isp-uv-es/hls-burn-scars-taco

HLS Burn Scars

This is a repackaging, not a new dataset. It is HLS Burn Scars by IBM-NASA Geospatial (Phillips, Roy, Ankur, Ramachandran), NASA / USGS HLS, 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:

@software{HLS_Foundation_2023,
  author = {Phillips, Christopher and Roy, Sujit and Ankur, Kumar and Ramachandran, Rahul},
  title  = {{HLS Foundation Burnscars Dataset}},
  month  = aug,
  year   = {2023},
  doi    = {10.57967/hf/0956},
  url    = {https://huggingface.co/ibm-nasa-geospatial/hls_burn_scars}
}

About the data

804 Harmonized Landsat Sentinel-2 (HLS S30) scenes of 512x512 px at 30 m over burn scars in the contiguous United States: six bands of surface reflectance with a binary burned-area mask. Converted from the original GeoTIFFs.

804 samples · splits: train 540 · validation 264 · tasks: semantic-segmentation

Packaged as TACO v3.

Full description

Splits. The official directory split, 540 train and 264 val. The release has no test split, and its own split is not MGRS-tile disjoint: 124 of 365 tiles appear on both sides, at different dates.

Radiometry. Reflectance is kept as float32 in physical units.

Nodata. The masks' -1 marks unlabelled pixels, and is rewritten as 255 and declared as the ignore index; the GeoTIFF headers claim nodata=-1000, which occurs nowhere in the data. The imagery's declared nodata of 0.0 is not propagated, because 0.0 is a real reflectance minimum in the blue band and honouring it would mask valid dark pixels.

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/hls-burn-scars-taco", "hls-burn-scars.zip", repo_type="dataset")
ds = Dataset(path)
plot_sample(ds[0])

or from a local copy:

ds = Dataset("hls-burn-scars.zip")
sample = ds[0]                      # {slot name: SlotValue}, arrays decoded
sample["hls"].array.shape

Metadata without decoding anything:

import taco
taco.read("hls-burn-scars.zip")              # one Arrow table, levels joined

Samples

What a sample contains

role	slot	holds	modality	detail
input	hls	raster	multispectral	6 band(s), unit 1, physical
target	burned	mask	label_raster	2 classes

Licence

CC-BY-4.0

Providers: IBM-NASA Geospatial (Phillips, Roy, Ankur, Ramachandran), NASA / USGS HLS

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.