Hierarchical Neural Coding for Controllable CAD Model Generation
This work addresses the problem of controllable CAD model generation for designers and engineers, representing an incremental advancement with a novel hierarchical approach.
The paper tackles the problem of generating controllable CAD models by introducing a hierarchical neural coding approach that represents design concepts as a three-level tree, enabling control over generation and completion. The method demonstrates superior performance on random generation tasks and enables novel interaction capabilities for conditional generation.
This paper presents a novel generative model for Computer Aided Design (CAD) that 1) represents high-level design concepts of a CAD model as a three-level hierarchical tree of neural codes, from global part arrangement down to local curve geometry; and 2) controls the generation or completion of CAD models by specifying the target design using a code tree. Concretely, a novel variant of a vector quantized VAE with "masked skip connection" extracts design variations as neural codebooks at three levels. Two-stage cascaded auto-regressive transformers learn to generate code trees from incomplete CAD models and then complete CAD models following the intended design. Extensive experiments demonstrate superior performance on conventional tasks such as random generation while enabling novel interaction capabilities on conditional generation tasks. The code is available at https://github.com/samxuxiang/hnc-cad.