CADForge: Agentic Single-View CAD Reconstruction with Explicit Geometry Reasoning

2026-10-03

AI Summary

This study addresses the challenges of incomplete geometric observations and complex part relationships in reconstructing editable parametric CAD models from single-view images. We propose CADForge, an agent-based framework that progressively translates a single image into a CadQuery program. By leveraging component decomposition and explicit geometric reasoning to derive modeling parameters, combined with iterative refinement via a review agent, the framework achieves high-fidelity reconstruction. Furthermore, it introduces a failure-guided toolkit construction mechanism to distill prior experience and maintains a compact parametric CAD memory bank for on-demand context retrieval, significantly enhancing robustness and efficiency. Experimental results demonstrate that our method consistently outperforms existing baselines in both reconstruction fidelity and perceptual quality across diverse single- and multi-part objects.

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.

Problem

• Single-view CAD reconstruction
• Parametric CAD models
• Editable CAD
• Geometric reasoning

Innovation

• Agentic Framework
• Explicit Geometry Reasoning
• CadQuery Program Synthesis
• Iterative Refinement
• Failure-guided Toolkit Construction