Search papers, labs, and topics across Lattice.
This paper introduces ExpConCAD, a novel framework for generating executable CAD programs from natural language descriptions by addressing the challenge of missing spatial constraints. By leveraging reusable design experience and understanding the underlying construction structure, ExpConCAD effectively infers and completes these constraints, leading to more accurate CAD generation. Extensive experiments validate the framework's effectiveness, highlighting the critical role of construction structure comprehension and experience memory in the process.
Missing spatial constraints in CAD generation can be effectively filled using past design experiences, significantly enhancing the accuracy of text-to-CAD translation.
Text-to-CAD aims to generate executable CAD programs from natural-language descriptions. However, real-world descriptions are often underspecified and omit critical spatial constraints required for valid CAD construction, a challenge that has been largely overlooked by existing methods. In this paper, we argue that missing spatial constraints should be inferred with respect to the underlying construction structure and informed by reusable design experience. Based on this insight, we propose ExpConCAD, an experience-enhanced framework for implicit spatial constraint completion. ExpConCAD first recovers the intended construction structure and constraint scopes, then retrieves relevant constraint-completion experience for similar scopes to complete the missing spatial constraints, and finally generates executable CadQuery programs. Extensive experiments demonstrate the effectiveness of ExpConCAD and provide insights into the role of construction structure understanding and experience memory in spatial constraint completion. Our code is available at: https://github.com/Hotjiashell/ExpConCAD.
NUS