Hierarchical Neural Coding for Controllable CAD Model Generation

Xiang Xu, Pradeep Kumar Jayaraman, Joseph George Lambourne, Karl D.D. Willis, Yasutaka Furukawa
Proceedings of the 40th International Conference on Machine Learning, PMLR 202:38443-38461, 2023.

Abstract

This paper presents 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. Concretely, 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 unconditional generation while enabling novel interaction capabilities on conditional generation tasks. The code is available at https://github.com/samxuxiang/hnc-cad.

Cite this Paper

BibTeX
@InProceedings{pmlr-v202-xu23f, title = {Hierarchical Neural Coding for Controllable {CAD} Model Generation}, author = {Xu, Xiang and Jayaraman, Pradeep Kumar and Lambourne, Joseph George and Willis, Karl D.D. and Furukawa, Yasutaka}, booktitle = {Proceedings of the 40th International Conference on Machine Learning}, pages = {38443--38461}, year = {2023}, editor = {Krause, Andreas and Brunskill, Emma and Cho, Kyunghyun and Engelhardt, Barbara and Sabato, Sivan and Scarlett, Jonathan}, volume = {202}, series = {Proceedings of Machine Learning Research}, month = {23--29 Jul}, publisher = {PMLR}, pdf = {https://proceedings.mlr.press/v202/xu23f/xu23f.pdf}, url = {https://proceedings.mlr.press/v202/xu23f.html}, abstract = {This paper presents 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. Concretely, 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 unconditional generation while enabling novel interaction capabilities on conditional generation tasks. The code is available at https://github.com/samxuxiang/hnc-cad.} }
Endnote
%0 Conference Paper %T Hierarchical Neural Coding for Controllable CAD Model Generation %A Xiang Xu %A Pradeep Kumar Jayaraman %A Joseph George Lambourne %A Karl D.D. Willis %A Yasutaka Furukawa %B Proceedings of the 40th International Conference on Machine Learning %C Proceedings of Machine Learning Research %D 2023 %E Andreas Krause %E Emma Brunskill %E Kyunghyun Cho %E Barbara Engelhardt %E Sivan Sabato %E Jonathan Scarlett %F pmlr-v202-xu23f %I PMLR %P 38443--38461 %U https://proceedings.mlr.press/v202/xu23f.html %V 202 %X This paper presents 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. Concretely, 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 unconditional generation while enabling novel interaction capabilities on conditional generation tasks. The code is available at https://github.com/samxuxiang/hnc-cad.
APA
Xu, X., Jayaraman, P.K., Lambourne, J.G., Willis, K.D. & Furukawa, Y.. (2023). Hierarchical Neural Coding for Controllable CAD Model Generation. Proceedings of the 40th International Conference on Machine Learning, in Proceedings of Machine Learning Research 202:38443-38461 Available from https://proceedings.mlr.press/v202/xu23f.html.

Related Material