isp-uv-es/avi-math-geomath-taco

AVI-Math GeoMath (training corpus)

This is a repackaging, not a new dataset. It is AVI-Math GeoMath (training corpus) by AVI-Math authors (VisionXLab), 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: apache-2.0

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{zhou2025avimath,
  author  = {Zhou, Yue and Feng, Litong and Lan, Mengcheng and Yang, Xue and Li, Qingyun and Ke, Yiping and Jiang, Xue and Zhang, Wayne},
  title   = {Multimodal Mathematical Reasoning Embedded in Aerial Vehicle Imagery: Benchmarking, Analysis, and Exploration},
  journal = {ISPRS Journal of Photogrammetry and Remote Sensing},
  volume  = {230},
  pages   = {289--303},
  year    = {2025},
  doi     = {10.1016/j.isprsjprs.2025.09.007}
}

About the data

215658 question/answer records over 22908 drone photographs: the machine-generated training corpus released with the AVI-Math benchmark, in the same question families -- recover the flight's above-ground level or pitch angle from the stated sensor parameters, count and price vehicles by colour and type, compute a frame's ground coverage.

22,908 samples · splits: train 20,121 · validation 2,787 · tasks: visual-question-answering

Packaged as TACO v3.

Full description

Records. Each record ships twice in the release, once with the worked step-by-step solution and once with the bare 'Answer: X'. Both are carried, as rec_response and rec_answer, and the two files pair question for question. The sample is the image, with its 6 to 32 records as parallel lists. Answers are machine-generated from the release's own scene annotations, not human-verified, and should not be treated as human-verified ground truth.

Imagery. Consumer-drone JPEGs at 4000x2250 stored byte-for-byte, no CRS and no published transfer function, so radiometry render and no approximate scale.

Splits. The release publishes no partition of this half. These 22,908 images are renderings of 502 photographs which are exactly AirSpatial's train frames, so the split follows AirSpatial's own train frames: validation holds whole source frames carved out of them, and there is no test part: AVI-Math's evaluation half is the avi-math container, whose images are renderings of AirSpatial's 120 test frames. The 360 images of avi-math are disjoint from these 22,908, so the two containers may be used together.

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

Metadata without decoding anything:

import taco
taco.read("avi-math-geomath/.tacocat")              # 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

apache-2.0

Providers: AVI-Math authors (VisionXLab)

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.