GenCAD-Self-Repairing: Feasibility Enhancement for 3D CAD Generation

TL;DR

GenCAD-Self-Repairing tackles the infeasibility of GenCAD-generated CAD programs by integrating diffusion-guided latent-space denoising with a self-repair pipeline. It introduces a Latent CAD Classifier and a Self-Supervised Learned Regressor, trained on a dedicated GenCAD-Self-Repairing dataset, to steer infeasible latent representations toward feasibility and to correct them before decoding. Empirical results show a notable improvement in feasibility (approximately 97% vs 93% for the baseline) and the ability to fix about two-thirds of previously infeasible designs, albeit with a modest degradation in geometric accuracy as measured by MMD. The approach expands usable training data for AI-driven CAD generation and holds practical potential for manufacturing, architecture, and product design by enabling higher-quality automated CAD generation from images.

Abstract

With the advancement of generative AI, research on its application to 3D model generation has gained traction, particularly in automating the creation of Computer-Aided Design (CAD) files from images. GenCAD is a notable model in this domain, leveraging an autoregressive transformer-based architecture with a contrastive learning framework to generate CAD programs. However, a major limitation of GenCAD is its inability to consistently produce feasible boundary representations (B-reps), with approximately 10% of generated designs being infeasible. To address this, we propose GenCAD-Self-Repairing, a framework that enhances the feasibility of generative CAD models through diffusion guidance and a self-repairing pipeline. This framework integrates a guided diffusion denoising process in the latent space and a regression-based correction mechanism to refine infeasible CAD command sequences while preserving geometric accuracy. Our approach successfully converted two-thirds of infeasible designs in the baseline method into feasible ones, significantly improving the feasibility rate while simultaneously maintaining a reasonable level of geometric accuracy between the point clouds of ground truth models and generated models. By significantly improving the feasibility rate of generating CAD models, our approach helps expand the availability of high-quality training data and enhances the applicability of AI-driven CAD generation in manufacturing, architecture, and product design.

Figures (11)

• Figure 1: GenCAD gencad architecture (Reproduced from gencad).
• Figure 2: GENCAD LIMITATION: Around 10% of CAD programs generated from GenCAD gencad result in invalid geometry kernels (Modified from gencad).
• Figure 3: Overview of GenCAD-Self-Repairing, which consists of a guided diffusion denoising process and a self-repair pipeline. The guided diffusion denoising directs latent vectors using classifier and regressor-based feedback, while the self-repair pipeline leverages an SSL Regressor to correct invalid latent representations. These mechanisms work together to improve the feasibility of CAD command generation.
• Figure 4: Creation process OF the DATASET.
• Figure 5: Five-image translation per B-rep: x(+), y(+), z(+), z(-), no change.