isp-uv-es/rescuenet-taco

RescueNet

This is a repackaging, not a new dataset. It is RescueNet by Rahnemoonfar, Chowdhury and Murphy (figshare collection 6647354), 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:

@article{rahnemoonfar2023rescuenet,
  title   = {RescueNet: A High Resolution UAV Semantic Segmentation Dataset for Natural Disaster Damage Assessment},
  author  = {Rahnemoonfar, Maryam and Chowdhury, Tashnim and Murphy, Robin},
  journal = {Scientific Data},
  volume  = {10},
  number  = {1},
  pages   = {913},
  year    = {2023},
  doi     = {10.1038/s41597-023-02799-4}
}

About the data

4494 high-resolution UAV frames of post-Hurricane-Michael Florida, each carrying both a pixel-level 11-class damage map and a whole-frame neighbourhood damage level.

4,494 samples · splits: test 450 · train 3,595 · validation 449 · tasks: semantic-segmentation, scene-classification

Packaged as TACO v3.

Full description

Legend. The segmentation legend grades buildings and roads by damage rather than by object class (Building_No_Damage through Building_Total_Destruction, Road-Clear against Road-Blocked), which is what makes it a damage-assessment benchmark. It has eleven entries although the release's own note says ten and then lists eleven; the enumeration is what the pixels hold, and a scan of all 4,494 masks found exactly the values 0..10 with every class occurring. Background (0) stays a class rather than an ignore label: nothing on disk distinguishes unannotated from none-of-the-other-ten.

One container. The masks and the classification tables cover identical frame sets in all three splits, checked both ways, and two containers would mean two copies of 12-megapixel JPEGs.

Imagery. Frames are JPEG and copied byte-for-byte, being lossy in with nothing to recover; masks are PNG and re-encoded as COGs at a measured 1.09x the PNG, which buys tiling on a 12 Mpx raster. Two frame sizes, 4000x3000 and 4056x3040, every mask header-checked against its own frame. UAV RGB with no geotransform and no published per-frame altitude: radiometry render, no CRS, no declared GSD.

Source. The authors' figshare release (Springer Nature collection 6647354, three clean zips, each extracting to exactly its own entry count), not the ResQAI/RescueNet Hub mirror, whose archive is incomplete.

Splits. Official, preserved: 3595 train, 449 val, 450 test.

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

Metadata without decoding anything:

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

Samples

What a sample contains

role	slot	holds	modality	detail
input	image	raster	optical	3 band(s), render
target	semantic	mask	label_raster	11 classes
target	damage	class_index		3 classes

Licence

CC-BY-4.0

Terms of use:

• Attribution required (CC-BY-4.0).

Required credits:

• RescueNet: Rahnemoonfar, Chowdhury and Murphy

Providers: Rahnemoonfar, Chowdhury and Murphy (figshare collection 6647354)

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.