isp-uv-es/dynamicearthnet-taco

DynamicEarthNet

This is a repackaging, not a new dataset. It is DynamicEarthNet by Planet Labs Inc., TUM (Toker et al.), 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:

@inproceedings{toker2022dynamicearthnet,
  title     = {DynamicEarthNet: Daily Multi-Spectral Satellite Dataset for Semantic Change Segmentation},
  author    = {Toker, Aysim and Kondmann, Lukas and Weber, Mark and Eisenberger, Marvin and Camero, Andr{\'e}s and Hu, Jingliang and Hoderlein, Ariadna Pregel and {\c{S}}enaras, {\c{C}}a{\u{g}}lar and Davis, Timothy and Cremers, Daniel and Marchisio, Giovanni and Zhu, Xiao Xiang and Leal-Taix{\'e}, Laura},
  booktitle = {Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR)},
  pages     = {21126--21135},
  year      = {2022},
  doi       = {10.1109/CVPR52688.2022.02048}
}

About the data

1,752 monthly samples over 73 AOIs worldwide, 2018-01 to 2019-12. Each sample is a daily Planet Fusion surface reflectance series of 28-31 frames at 3 m (4 bands: blue, green, red, NIR; reflectance x 1e4) against one 7-class land-cover raster for that month.

1,752 samples · splits: test 240 · train 1,272 · validation 240 · tasks: semantic-segmentation, change-detection

Packaged as TACO v3.

Full description

Label. It ships as seven one-hot bands and is argmaxed here. It is not strictly one-hot: 0.20% of pixels carry more than one class and 0.0003% none, so multi_class_frac and unlabelled_frac are recorded per sample, and an eighth index unlabelled carries the empty pixels. The native legend maps one-to-one onto ml:lc11.

Omitted layers. The release's Sentinel-1 (8 bands), Sentinel-2 (12 bands) and Planet Fusion QA (9 bands) layers are not carried. All three ship without any band description and the release documents none, so their channel identity cannot be declared; 86 and 36 of the samples respectively have no such file at all.

Splits. The release's own AOI-disjoint split: 1272 train, 240 val, 240 test.

Getting started

git clone https://github.com/OscarPellicer/taco
pip install -e "taco/python[ml]"     # builds the native reader: needs CMake and Ninja

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

Metadata without decoding anything:

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

Samples

What a sample contains

role	slot	holds	modality	detail
input	sr	raster_series	optical	4 band(s), unit 1, scaled
target	lc	mask	label_raster	8 classes

Licence

CC-BY-SA-4.0

Terms of use:

• Adaptations must be shared under CC-BY-SA-4.0.
• Credit the original authors and the data sources listed below.

Required credits:

• DynamicEarthNet (Toker et al., 2022), Technical University of Munich and Planet Labs
• © 2021 Planet Labs Inc.

Providers: Planet Labs Inc., TUM (Toker et al.)

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.