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

Ahmet Serdar Karadeniz, Dimitrios Mallis, Nesryne Mejri, Kseniya Cherenkova, Anis Kacem, Djamila Aouada

🎉 This work is accepted for publication in BMVC2024!!

Download CPTSketchGraphs dataset (17.1 GB).

This is the repository for the CPTSketchGraphs dataset, a collection of 80 million Constraint Preserving Transformations (CPTs) of CAD sketches. CPTs are derived from CAD sketches of the large-scale SketchGraphs dataset. We aim for CPTSketchGraphs to serve as a valuable resource for advancing research in the CAD sketch domain.

Constraint Preserving Transformations (CPTs)

Our proposed Constraint Preserving Transformations (CPTs), is a novel augmentation strategy particularly tailored for CAD sketches. CPTs leverage the existing geometric constraints within CAD sketches to generate plausible variations of the original design.

A random local perturbation, such as the translation of a sketch point, is automatically applied to the CAD sketch. Due to the constraints associated with the manipulated point, this local change cascades across the sketch, modifying the parameterization of all connected primitives. The augmentation process is enabled via integration with the FreeCAD Python API.

CPTSketchGraphs Dataset

You can access the dataset using this SharePoint link. The cptsketchgraphs.npy(18 GB) file should be downloaded within the CPTSketchGraphs/data directory.

The CPTs shared in this work are derived from CAD sketches from the SketchGraphs dataset. We use the preprocessing as in Vitruvion. For effective visualization of both CPTs and original CAD sketches, the user can optionally download the filtered sequences in sg_filtered_unique.npy as described here.

You can visualize CPTs using our demo notebook.

Setup

The environment for runnin the demo notebook can be created as follows:

conda create --name cptsketchgraphs python=3.10
pip install -r requirements.txt

Citation

If you use this dataset in your research, please cite:

@inproceedings{davinci2024,
    title={DAVINCI: A Single-Stage Architecture for Constrained CAD Sketch Inference},
    author={Ahmet Serdar Karadeniz and Dimitrios Mallis and Nesryne Mejri and Kseniya Cherenkova and Anis Kacem and Djamila Aouada},
    booktitle={British Machine Vision Conference},
    year={2024}
  }