ConCAD improves image to 3D CAD model conversion with better constraint awareness
ConCAD: Constraint-Aware Image-to-CAD Generation with Dual-Granularity Rewards
Computer Vision and Pattern Recognition
Summary
Turning pictures into 3D computer-aided design (CAD) models is tricky, especially when models that look similar may have very different internal design rules. The authors present ConCAD, which uses two types of rewards to better capture both the shape and the underlying design constraints from an image. This approach helps distinguish between models that occupy the same space but have different design structures. They also introduce a way to measure how well these geometric constraints are recovered, showing their method improves both shape accuracy and design intent.
What this means in practice
- •For cad software developers: Enhance CAD programs to generate more accurate parametric models from images, preserving both geometry and design constraints.$Commercial implications: Enables improved image-based CAD modeling tools for designers requiring valid parametric outputs, which can be sold to engineering and design professionals.
- •For industrial designers: Use images of objects to generate editable 3D CAD models that faithfully represent physical shape and design intent for prototype development.
Authors
Chenxi Zhai, Xi Cheng, Hang Cheng, Zhicheng Guan, Mingyu Fan, Yanzhe Tang, Pingfa Feng, Long Zeng
Abstract
Image-to-CAD generation seeks executable parametric programs that recover both the geometry and design intent of a reference object. Existing systems are commonly evaluated by validity and shape overlap, although two solids with similar volume can encode different CAD relations. We introduce ConCAD, a constraint-aware image-to-CAD framework optimized via Group Relative Policy Optimization (GRPO) with rewards at two complementary granularities: a code-level constraint reward and an execution-level geometric reward. This complementary design disambiguates structurally distinct yet volumetrically similar shapes while ensuring valid 3D geometry. To verify that these rewards recover geometry and design intent, we introduce a B-rep geometric constraint satisfaction rate (G-CSR), which analytically extracts and evaluates geometric constraints from boundary representations. Experiments on the DeepCAD and Zero2CAD demonstrate that ConCAD achieves the best IoU and Chamfer Distance over competitive baselines, while also outperforming them on G-CSR, validating its superior recovery of both geometric fidelity and parametric design intent.