DeepCAD (dataset)
178K sketch-and-extrude sequences parsed from Onshape, the foundational corpus nearly every text-to-CAD benchmark builds on.
What it does and how it connects
DeepCAD is the dataset and code for the ICCV 2021 paper of the same name, by Rundi Wu, Chang Xiao and Changxi Zheng. The data is 178K sketch-and-extrude sequences parsed from Onshape public documents, with links from the ABC dataset, and the repository also releases the Onshape CAD data parsing scripts.
The repository provides the paper's generative model code: an autoencoder trained on CAD sequences plus a latent GAN for random generation, with evaluation by command accuracy, parameter accuracy, chamfer distance, invalid ratio, coverage, MMD and JSD. Pre-trained models are hosted on the Columbia project page.
The README also provides scripts to visualize CAD models and export results to STEP files, which load in almost all modern CAD software. The dataset is the foundational corpus that later text-to-CAD benchmarks, including Text2CAD and CADBench, build on.
The catalog repo mtfgqd/deepcad is a fork of rundiwu/DeepCAD (created 2026-05-29, 0 stars); the stats reflect the fork, not the upstream repo.
Alternatives
Compare side by side- Text2CAD
A framework and dataset of around 170K models that generates sketch-and-extrude sequences from text, rather than a dataset of sketch-and-extrude sequences parsed from Onshape.
- CADBench (Doris et al.)
A unified 18,000-sample benchmark turning mesh, image or render inputs into STEP or mesh, rather than a dataset of Onshape sketch-and-extrude sequences.
- Text2CAD-Bench
Tests text-to-parametric-CAD on 600 human-curated examples across four complexity levels, rather than a dataset of sketch-and-extrude sequences.
- BenchCAD
Scores execution-verified CadQuery programs by deterministic voxel IoU across 106 industrial part families, rather than a dataset of sketch-and-extrude sequences.
- NURBGen
Generates high-fidelity CAD models from text using NURBS surfaces, rather than a dataset of sketch-and-extrude sequences.
Sources
Not verified
- Pricing: none stated on the repo (open source research code and data).
- Real user voices: no real linked post read for this page.
- No GitHub releases exist, so latest release fields are null.
- Related Error Index entries: no matching entry verified.
- The catalog repo is a fork (mtfgqd/deepcad, created 2026-05-29, 0 stars); the original upstream DeepCAD repository was not the catalog URL.
Last verified November 6, 2026. Listings are informational. The evidence score measures public, verifiable evidence and is not an endorsement or a compatibility claim.