abc-step-1000#

The first 1000 STEP files from the ABC dataset, hosted on GitHub Pages for convenient direct download.

• Browse: https://concept-collection.github.io/abc-step-1000/
• Manifest: https://concept-collection.github.io/abc-step-1000/index.json

The files are served gzip-compressed (.step.gz, ~300 MB total; ~1.6 GB uncompressed). index.json lists every file with its download path, model ID, and compressed/uncompressed sizes.

Downloading#

A single file:

curl -sL https://concept-collection.github.io/abc-step-1000/step/00000002_1ffb81a71e5b402e966b9341_step_001.step.gz \
  | gunzip > model.step

All files, using the manifest:

BASE=https://concept-collection.github.io/abc-step-1000
curl -sL $BASE/index.json | jq -r '.files[].path' \
  | xargs -P 8 -I{} sh -c 'curl -sL "$1/$2" | gunzip > "$(basename "$2" .gz)"' _ $BASE {}

In the browser, decompress with DecompressionStream:

const res = await fetch(url);
const step = await new Response(
  res.body.pipeThrough(new DecompressionStream('gzip'))
).text();

How it is built#

The GitHub Actions workflow (deploy.yml) downloads the first chunk of the STEP format (abc_0000_step_v00.7z, ~1.6 GB), extracts the first 1000 model directories, gzips each .step file, generates index.json, and deploys the result to GitHub Pages. The content is a fixed slice of the dataset, so the workflow runs on manual dispatch only.

The canonical host (archive.nyu.edu) rate-limits downloads per IP and often serves CI runners an HTML restrictions page instead of the archive, so the build validates the download and falls back to a release asset on this repository that repacks just the first 1000 model directories of the canonical chunk (byte-identical files).

Source and acknowledgments#

All CAD models come from the ABC dataset:

Koch, Sebastian and Matveev, Albert and Jiang, Zhongshi and Williams, Francis and Artemov, Alexey and Burnaev, Evgeny and Alexa, Marc and Zorin, Denis and Panozzo, Daniele. ABC: A Big CAD Model Dataset For Geometric Deep Learning. CVPR 2019.

@InProceedings{Koch_2019_CVPR,
  author = {Koch, Sebastian and Matveev, Albert and Jiang, Zhongshi and Williams, Francis and Artemov, Alexey and Burnaev, Evgeny and Alexa, Marc and Zorin, Denis and Panozzo, Daniele},
  title = {ABC: A Big CAD Model Dataset For Geometric Deep Learning},
  booktitle = {The IEEE Conference on Computer Vision and Pattern Recognition (CVPR)},
  month = {June},
  year = {2019}
}

Please cite the paper if you use these models. The ABC dataset authors are grateful to Onshape for providing the CAD models and support.

The copyright of the CAD models is owned by their creators; for licensing details see the Onshape Terms of Use 1.g.ii. The dataset authors give no warranties regarding the dataset.