CADEvolve: Creating Realistic CAD via Program Evolution

TL;DR

CADEvolve introduces an evolution-based method and dataset for generating complex, realistic CAD programs from simple primitives, leveraging VLM-guided edits to achieve industrial-grade designs and state-of-the-art results.

Contribution

It presents a novel evolution pipeline and a large dataset for training models to generate complex CAD programs, overcoming data limitations and grounding challenges.

Findings

Generated 8k complex CAD parts as parametric scripts.

Created a dataset of 1.3 million scripts with geometry data.

Achieved state-of-the-art results on multiple CAD benchmarks.

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.