isp-uv-es/rsvqa-lr-taco

RSVQA-LR (2k evaluation subset)

This is a repackaging, not a new dataset. It is RSVQA-LR (2k evaluation subset) by Sylvain Lobry et al. (RSVQA), Copernicus / Sentinel-2, 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: ODbL-1.0, 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{lobry2020rsvqa,
  title   = {RSVQA: Visual Question Answering for Remote Sensing Data},
  author  = {Lobry, Sylvain and Marcos, Diego and Murray, Jesse and Tuia, Devis},
  journal = {IEEE Transactions on Geoscience and Remote Sensing},
  volume  = {58},
  number  = {12},
  pages   = {8555--8566},
  year    = {2020}
}

About the data

2000 closed-form question/answer pairs over 100 Sentinel-2 tiles of 256x256 px at 10 m: presence, counting, comparison and rural/urban questions from the 2020 benchmark that defined remote-sensing VQA.

100 samples · splits: validation 100 · tasks: visual-question-answering

Packaged as TACO v3.

Full description

Scope. A 2,000-row evaluation subset of the release's validation split, not the ~77k-pair corpus. The full RSVQA-LR is distributed by its authors and is not included here.

Layout. Samples are grouped by image: each tile is stored once with its 20 questions as parallel qa_question and qa_answer lists.

• The mirror carries no filenames, so image_id is the SHA-1 of the encoded image bytes.
• The mirror carries no question-type column, so the release's presence/count/comparison/rural-urban taxonomy is absent rather than inferred from the question text.

Imagery. PNGs are stored byte-for-byte. There is no CRS or transform, so radiometry is RENDER.

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/rsvqa-lr-taco", "rsvqa-lr.zip", repo_type="dataset")
ds = Dataset(path)
plot_sample(ds[0])

or from a local copy:

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

Metadata without decoding anything:

import taco
taco.read("rsvqa-lr.zip")              # one Arrow table, levels joined

Samples

What a sample contains

role	slot	holds	modality	detail
input	image	raster	optical	3 band(s), render
input	questions	text_sequence	text
target	answers	text_sequence	text

Licence

ODbL-1.0, CC-BY-4.0

Terms of use:

• The questions and answers are derived from OpenStreetMap; a database adapted from them may be shared only under ODbL-1.0.
• The dataset is otherwise licensed under CC-BY-4.0.

Required credits:

• RSVQA: Lobry, Marcos, Murray and Tuia
• © OpenStreetMap contributors (ODbL)
• Contains modified Copernicus Sentinel data

Providers: Sylvain Lobry et al. (RSVQA), Copernicus / Sentinel-2

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.