Cadforge builds editable 3d cad models from single images
CADForge: Agentic Single-View CAD Reconstruction with Explicit Geometry Reasoning
Artificial Intelligence
Summary
Creating 3D editable CAD models from just one picture is hard because you can't see the whole object and parts connect in complex ways. The authors developed CADForge, a step-by-step system that breaks down an object into meaningful parts and figures out the exact shapes and rules for each part using math. It then writes code that builds the 3D model piece by piece, checking and improving itself as it goes. CADForge learns from its mistakes and remembers previous work to get better and faster. Tests show it does better than older methods at making accurate and usable CAD models from one image.
What this means in practice
- •For industrial designers: Convert a photo of a product into an editable CAD model to speed up design modifications.
- •For automotive part manufacturers: Generate parametric CAD files from single images of mechanical parts to support rapid prototyping and replacement part creation.
Authors
Keyang Lu, Zhifei Yang, Tianao Dong, Mingzhe Xing, Zhen Xiao, Yikai Wang
Abstract
Reconstructing editable parametric CAD models from a single-view image is of great practical value for modern manufacturing, yet remains challenging due to incomplete geometric observations and complex inter-part relationships. To address it, we propose CADForge, an agentic framework that progressively converts a single image into CadQuery programs. CADForge decomposes an object into CAD-meaningful components and performs explicit geometric reasoning for each component, a process that first identifies CAD-relevant constraints and then translates them into precise modeling parameters through mathematical code. The inferred parameters then drive component-wise synthesis of executable CadQuery programs, with a review agent evaluating the resulting geometry and providing targeted feedback for iterative refinement. To further improve robustness and efficiency, CADForge incorporates a failure-guided toolkit construction mechanism to distill accumulated experience into tools, and maintains a compact parametric CAD memory for retrieving modeling context on demand. Experiments on diverse single- and multi-part objects show that CADForge consistently outperforms existing baselines in reconstruction fidelity and perceptual quality, demonstrating an effective approach to accurate single-view CAD reconstruction.