DAVINCI: A Single-Stage Architecture for Constrained CAD Sketch Inference

TL;DR

DAVINCI presents a novel single-stage transformer that jointly infers CAD sketch primitives and their geometric constraints directly from raster images, reducing error propagation typical of two-stage pipelines. The architecture uses prim- and constr-embeddings to produce a sketch graph and leverages Hungarian matching for alignment, achieving state-of-the-art results on the SketchGraphs dataset for both precise and hand-drawn sketches. A key contribution is Constraint-Preserving Transformations (CPTs), an augmentation strategy that preserves design constraints while expanding data diversity, enabling strong performance even with only 0.1% of the original data and yielding the CPTSketchGraphs dataset with about 80 million sketches. The work demonstrates significant gains in constraint inference when training jointly with primitive parameterization and provides a practical path toward data-efficient constrained CAD sketch understanding and reverse engineering applications, including cross-section analysis.

Abstract

This work presents DAVINCI, a unified architecture for single-stage Computer-Aided Design (CAD) sketch parameterization and constraint inference directly from raster sketch images. By jointly learning both outputs, DAVINCI minimizes error accumulation and enhances the performance of constrained CAD sketch inference. Notably, DAVINCI achieves state-of-the-art results on the large-scale SketchGraphs dataset, demonstrating effectiveness on both precise and hand-drawn raster CAD sketches. To reduce DAVINCI's reliance on large-scale annotated datasets, we explore the efficacy of CAD sketch augmentations. We introduce Constraint-Preserving Transformations (CPTs), i.e. random permutations of the parametric primitives of a CAD sketch that preserve its constraints. This data augmentation strategy allows DAVINCI to achieve reasonable performance when trained with only 0.1% of the SketchGraphs dataset. Furthermore, this work contributes a new version of SketchGraphs, augmented with CPTs. The newly introduced CPTSketchGraphs dataset includes 80 million CPT-augmented sketches, thus providing a rich resource for future research in the CAD sketch domain.

Figures (7)

• Figure 1: We propose DAVINCI, a novel single-stage network for constrained CAD sketch parameterization. DAVINCI effectively parameterizes different types of input sketches, from precise to hand-drawn as well as 2D cross-sections. We also introduce Constraint-Preserving Transformations (CPTs), i.e. an augmentation strategy tailored to constrained CAD sketches.
• Figure 2: Overview of the DAVINCI architecture. A raster CAD sketch image is processed by a transformer encoder/decoder. Predicted [prim]-embeddings and [constr]-embeddings are used to infer primitive tokens and geometric constraints between primitives.
• Figure 3: Example Constraint-Preserving Transformations (CPTs) of CAD sketches from CPTSketchGraphs. CPTs are generated via intergration with the FreeCAD API FreeCAD.
• Figure 4: Formation of a CPT via random local perturbations.
• Figure 5: Qualitative comparison with Vitruvion seff2022vitruvion.