We represent cad design features as three-level code tree and use it for controllable generation.
We present a novel generative model for Computer Aided Design (CAD) that 1) represents high-level design concepts of a CAD model as a three-level hierarchical tree of neural codes, from global part arrangement down to local curve geometry; and 2) controls the generation or completion of CAD models by specifying the target design using a code tree.
A novel variant of a vector quantized VAE with "masked skip connection" extracts design variations as neural codebooks at three levels. Two-stage cascaded auto-regressive transformers learn to generate code trees from incomplete CAD models and then complete CAD models following the intended design. Extensive experiments demonstrate superior performance on conventional tasks such as random generation while enabling novel interaction capabilities on conditional generation tasks.
Given a code tree, user can edit the code nodes and achieve local and global modifications across the CAD hierarchy.
Our code tree generator can predict a set of likely codes from user partial input and autocomplete the full CAD model.
With the code tree fixed, user can preserve the current design while making local edits to the cad model parameters to iteratively refine it.
@article{xu2023hierarchical,
title = {Hierarchical Neural Coding for Controllable CAD Model Generation},
author = {Xu, Xiang and Jayaraman, Pradeep Kumar and Lambourne, Joseph G and Willis, Karl DD and Furukawa, Yasutaka},
journal = {arXiv preprint arXiv:2307.00149},
year = {2023}
}
This research is partially supported by NSERC Discovery Grants with Accelerator Supplements and DND/NSERC Discovery Grant Supplement, NSERC Alliance Grants, and John R. Evans Leaders Fund (JELF).