Revisiting CAD Model Generation by Learning Raster Sketch

2025-03-02

AI Summary

Traditional CAD model generation methods suffer from limited flexibility in curve representation and poor user controllability. To address this, we propose the first end-to-end generative framework that jointly models rasterized sketches and extrusion operations. Our method replaces discrete parametric curve sequences with raster sketches and introduces a two-stage conditional diffusion model that jointly synthesizes high-fidelity extruded solids and their corresponding sketches. Crucially, the framework enables latent-space continuous interpolation and interactive sketch editing. Evaluated on three tasks—unconditional generation, conditional generation, and sketch-guided editing—our approach achieves state-of-the-art performance in all, significantly improving geometric plausibility, generation fidelity, and intuitive user control. By unifying sketch-based input with parametric solid modeling in a learnable generative pipeline, our work establishes a novel paradigm for automated CAD modeling.

Technology Category

• Computer Vision: Diffusion Models for Vision
• Machine Learning: Deep Generative Models & Autoencoders
• Natural Language Processing: Generation

Application Category

• User Modeling, Personalization and Recommendation: Explainable and interpretable methods for personalization
• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs
• Graph Algorithms and Modeling for the Web: Foundation models and LLMs for Web-related graphs

Abstract

The integration of deep generative networks into generating Computer-Aided Design (CAD) models has garnered increasing attention over recent years. Traditional methods often rely on discrete sequences of parametric line/curve segments to represent sketches. Differently, we introduce RECAD, a novel framework that generates Raster sketches and 3D Extrusions for CAD models. Representing sketches as raster images offers several advantages over discrete sequences: 1) it breaks the limitations on the types and numbers of lines/curves, providing enhanced geometric representation capabilities; 2) it enables interpolation within a continuous latent space; and 3) it allows for more intuitive user control over the output. Technically, RECAD employs two diffusion networks: the first network generates extrusion boxes conditioned on the number and types of extrusions, while the second network produces sketch images conditioned on these extrusion boxes. By combining these two networks, RECAD effectively generates sketch-and-extrude CAD models, offering a more robust and intuitive approach to CAD model generation. Experimental results indicate that RECAD achieves strong performance in unconditional generation, while also demonstrating effectiveness in conditional generation and output editing.

Problem

• Generates CAD models using raster sketches and 3D extrusions.
• Overcomes limitations of traditional parametric line/curve representations.
• Enhances user control and geometric representation in CAD generation.

Innovation

• RECAD generates raster sketches for CAD models.
• Uses two diffusion networks for extrusion and sketches.
• Enables continuous latent space interpolation for CAD.