CADEvolve: Creating Realistic CAD via Program Evolution

2026-02-18

AI Summary

This work addresses the limitations of existing public CAD datasets, which often lack complex operations, multi-step combinations, and explicit design intent, thereby hindering the generalization of AI models in real-world industrial settings. To overcome this, we propose a novel generation framework that integrates program evolution with vision-language model (VLM) guidance. Starting from simple geometric primitives, our approach iteratively constructs executable CadQuery programs of industrial-level complexity through VLM-driven editing and validation. This is the first method to combine program evolution with VLM-based reasoning to produce a high-quality, parameterized CAD dataset encompassing the full CadQuery operation set, enhanced by multi-stage post-processing and data augmentation. The resulting dataset comprises 1.3 million scripts and achieves state-of-the-art performance on established Image2CAD benchmarks, including DeepCAD, Fusion 360, and MCB.

Technology Category

• Computer Vision: Diffusion Models for Vision
• Natural Language Processing: Code Generation / Program Synthesis from Natural Language
• Humans and AI: Game Design — Procedural Content Generation & Storytelling

Application Category

• Semantics and Knowledge: Data modeling to support human-machine intelligence, including LLMs agents, intelligent system behavior, explanations, and user-friendly interactions
• Economics, Online Markets and Human Computation: Cost models of using LLMs in production systems
• Graph Algorithms and Modeling for the Web: Foundation models and LLMs for Web-related graphs

Abstract

Computer-Aided Design (CAD) delivers rapid, editable modeling for engineering and manufacturing. Recent AI progress now makes full automation feasible for various CAD tasks. However, progress is bottlenecked by data: public corpora mostly contain sketch-extrude sequences, lack complex operations, multi-operation composition and design intent, and thus hinder effective fine-tuning. Attempts to bypass this with frozen VLMs often yield simple or invalid programs due to limited 3D grounding in current foundation models. We present CADEvolve, an evolution-based pipeline and dataset that starts from simple primitives and, via VLM-guided edits and validations, incrementally grows CAD programs toward industrial-grade complexity. The result is 8k complex parts expressed as executable CadQuery parametric generators. After multi-stage post-processing and augmentation, we obtain a unified dataset of 1.3m scripts paired with rendered geometry and exercising the full CadQuery operation set. A VLM fine-tuned on CADEvolve achieves state-of-the-art results on the Image2CAD task across the DeepCAD, Fusion 360, and MCB benchmarks.

Problem

• CAD dataset
• complex operations
• design intent
• 3D grounding
• program generation

Innovation

• program evolution
• CAD generation
• visual language model
• parametric modeling
• dataset synthesis