Regularized Diffusion Modeling for CAD Representation Generation

Yunpeng Bai, Ruisheng Wang, Qixing Huang

Keywords: CAD, Diffusion Regularization, 3D Generation

Abstract: Computer-Aided Design (CAD) has significant practical value in various industrial applications. However, achieving high-quality and diverse shape generation, as well as flexible conditional control, remains a challenge in the field of CAD model generation. To address these issues, we propose CADiffusion, a diffusion-based generative model with a hierarchical latent representation tailored to the complexities of CAD design processes. To enhance the performance and reliability of the model in generating accurate CAD models, we have developed a specialized decoder with regularization strategies that navigate through the noise space of the diffusion model, smoothing the results. This approach not only improves the diversity and quality of the generated CAD models but also enhances their practical applicability, marking a significant advancement in the integration of generative models and automated CAD systems.

Primary Area: generative models

Code Of Ethics: I acknowledge that I and all co-authors of this work have read and commit to adhering to the ICLR Code of Ethics.

Submission Guidelines: I certify that this submission complies with the submission instructions as described on https://iclr.cc/Conferences/2025/AuthorGuide.

Reciprocal Reviewing: I understand the reciprocal reviewing requirement as described on https://iclr.cc/Conferences/2025/CallForPapers. If none of the authors are registered as a reviewer, it may result in a desk rejection at the discretion of the program chairs. To request an exception, please complete this form at https://forms.gle/Huojr6VjkFxiQsUp6.

Anonymous Url: I certify that there is no URL (e.g., github page) that could be used to find authors’ identity.

No Acknowledgement Section: I certify that there is no acknowledgement section in this submission for double blind review.

Submission Number: 1972