TLDR: We formulate the task of discovering modular CAD sketch concepts as program library induction and propose a self-supervised deep learning framework that discovers modular libraries with end-to-end training

Sketch design concepts are recurring patterns found in parametric CAD sketches. Though rarely explicitly formalized by the CAD designers, these concepts are implicitly used in design for modularity and regularity. In this paper, we propose a learning based approach that discovers the modular concepts by induction over raw sketches. We propose the dual implicit-explicit representation of concept structures that allows implicit detection and explicit generation, and the separation of structure generation and parameter instantiation for parameterized concept generation, to learn modular concepts by end-to-end training. We demonstrate the design concept learning on a large scale CAD sketch dataset and show its applications for design intent interpretation and auto-completion.

Framework illustration. (a) The detection module is a transformer network that detects from the sketch sequence \( [t^0_i] \) implicitly encoded concepts \( [\mathbf{q}_i] \) and their composition \( \mathbf{q}_R \). (b) Each \( \mathbf{q} \) is quantized against the concept library \( \mathbf{L}^1 \) to obtain prototype \( \mathbf{q}' \), which is expanded by the structure network into an explicit structure \( \mathbf{T}^1 \) and further instantiated by the parameter network into \( t^1 \). (c) The collection of \( [t^1_i] \) are assembled by the composition operator \( R_S \) generated from \( \mathbf{q}_R \) to obtain the final generated sketch graph, which is compared with the input sketch for loss computation.

          @inproceedings{yang2022sketchconcept,
          author = {Yuezhi Yang, Hao Pan},
          booktitle = {Advances in Neural Information Processing Systems},
          title = {Discovering Design Concepts for CAD Sketches},
          volume = {35},
          year = {2022}
          }