isp-uv-es/spacenet3-taco

SpaceNet 3: Road Network Extraction

This is a repackaging, not a new dataset. It is SpaceNet 3: Road Network Extraction by SpaceNet LLC / CosmiQ Works, Maxar (WorldView-3), 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-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{vanetten2018spacenet,
  title   = {SpaceNet: A Remote Sensing Dataset and Challenge Series},
  author  = {Van Etten, Adam and Lindenbaum, Dave and Bacastow, Todd M.},
  journal = {arXiv:1807.01232},
  year    = {2018}
}

About the data

2544 WorldView-3 chips over Las Vegas, Paris, Shanghai and Khartoum. Each carries a 1300x1300 panchromatic band at 0.31 m, the co-located 325x325 8-band multispectral cube at 1.24 m, and the road centrelines through it with SpaceNet's inferred speed per segment.

2,544 samples · splits: test 550 · train 1,449 · validation 545 · tasks: semantic-segmentation

Packaged as TACO v3.

Full description

Target. Three views of the label:

• A 2 m buffered road mask.
• The polylines themselves, so the graph survives.
• A raster of inferred speed over the same buffer.

Products. The pan-sharpened products are not carried: they are derived from the two that are.

Splits. The challenge test chips ship without labels, so the split here is not the publisher's: it is cut by longitude within each city. Train 1449, val 545, test 550, after 5 labelled chips with no valid pixel in either product are dropped.

Held-out half

spacenet3_test holds the 925 samples whose targets the publisher withheld. Same inputs, no target slots: it is there to be predicted on and submitted, and it belongs in no training mixture.

import os
from huggingface_hub import snapshot_download
from taco.ml import Dataset

root = snapshot_download("isp-uv-es/spacenet3-taco", repo_type="dataset",
                         allow_patterns=["spacenet3_test", "spacenet3_test/*", "spacenet3_test/.tacocat/*"])
ds = Dataset(os.path.join(root, "spacenet3_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/spacenet3-taco", repo_type="dataset", ignore_patterns=["spacenet3_test", "spacenet3_test/*"])
ds = Dataset(path)                  # the repository is the container: parts + .tacocat
plot_sample(ds[0])

or from a local copy:

ds = Dataset("spacenet3")
sample = ds[0]                      # {slot name: SlotValue}, arrays decoded
sample["pan"].array.shape

Metadata without decoding anything:

import taco
taco.read("spacenet3/.tacocat")              # one Arrow table, levels joined

Samples

What a sample contains

role	slot	holds	modality	detail
input	pan	raster	optical	1 band(s), digital_number
input	ms	raster	multispectral	8 band(s), digital_number
target	roads	mask	label_raster	2 classes
target	centrelines	polygon		unit degree

Licence

CC-BY-SA-4.0

Terms of use:

• Adapted material must be shared under CC-BY-SA-4.0.

Required credits:

• The SpaceNet Dataset by SpaceNet Partners is licensed under CC-BY-SA-4.0
• Imagery: Maxar WorldView-3

Providers: SpaceNet LLC / CosmiQ Works, Maxar (WorldView-3)

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.