3The first 1000 STEP files from the [ABC dataset](https://deep-geometry.github.io/abc-dataset/),
4hosted on GitHub Pages for convenient direct download.
6- **Browse:** https://concept-collection.github.io/abc-step-1000/
7- **Manifest:** https://concept-collection.github.io/abc-step-1000/index.json
9The files are served gzip-compressed (`.step.gz`, ~300 MB total; ~1.6 GB
10uncompressed). `index.json` lists every file with its download path, model ID,
11and compressed/uncompressed sizes.
13## Downloading
15A single file:
17```sh
18curl -sL https://concept-collection.github.io/abc-step-1000/step/00000002_1ffb81a71e5b402e966b9341_step_001.step.gz \
19 | gunzip > model.step
20```
22All files, using the manifest:
24```sh
25BASE=https://concept-collection.github.io/abc-step-1000
26curl -sL $BASE/index.json | jq -r '.files[].path' \
27 | xargs -P 8 -I{} sh -c 'curl -sL "$1/$2" | gunzip > "$(basename "$2" .gz)"' _ $BASE {}
28```
30In the browser, decompress with
31[`DecompressionStream`](https://developer.mozilla.org/en-US/docs/Web/API/DecompressionStream):
33```js
34const res = await fetch(url);
35const step = await new Response(
36 res.body.pipeThrough(new DecompressionStream('gzip'))
37).text();
38```
40## How it is built
42The GitHub Actions workflow ([deploy.yml](.github/workflows/deploy.yml))
43downloads the first chunk of the STEP format (`abc_0000_step_v00.7z`, ~1.6 GB),
44extracts the first 1000 model directories, gzips each `.step` file, generates
45`index.json`, and deploys the result to GitHub Pages. The content is a fixed
46slice of the dataset, so the workflow runs on manual dispatch only.
48## Source and acknowledgments
50All CAD models come from the **ABC dataset**:
52> Koch, Sebastian and Matveev, Albert and Jiang, Zhongshi and Williams, Francis
53> and Artemov, Alexey and Burnaev, Evgeny and Alexa, Marc and Zorin, Denis and
54> Panozzo, Daniele. *ABC: A Big CAD Model Dataset For Geometric Deep Learning.*
55> CVPR 2019.
57```bibtex
58@InProceedings{Koch_2019_CVPR,
59 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},
60 title = {ABC: A Big CAD Model Dataset For Geometric Deep Learning},
61 booktitle = {The IEEE Conference on Computer Vision and Pattern Recognition (CVPR)},
62 month = {June},
63 year = {2019}
64}
65```
67Please cite the paper if you use these models. The ABC dataset authors are
68grateful to [Onshape](https://www.onshape.com/) for providing the CAD models
69and support.
71The copyright of the CAD models is owned by their creators; for licensing
72details see the
73[Onshape Terms of Use 1.g.ii](https://www.onshape.com/en/legal/terms-of-use#your_content).
74The dataset authors give no warranties regarding the dataset.