NeuroNURBS: Learning Efficient Surface Representations for 3D Solids
This work addresses efficiency and quality issues in 3D solid representation for Computer-Aided Design, offering incremental improvements over existing methods.
The paper tackles the inefficiency and lack of precision in representing 3D solids using UV-grid approximations by proposing NeuroNURBS, a method that directly encodes NURBS surface parameters, resulting in reduced GPU consumption by 86.7% and memory usage by 79.9%, while improving FID scores in solid generation.
Boundary Representation (B-Rep) is the de facto representation of 3D solids in Computer-Aided Design (CAD). B-Rep solids are defined with a set of NURBS (Non-Uniform Rational B-Splines) surfaces forming a closed volume. To represent a surface, current works often employ the UV-grid approximation, i.e., sample points uniformly on the surface. However, the UV-grid method is not efficient in surface representation and sometimes lacks precision and regularity. In this work, we propose NeuroNURBS, a representation learning method to directly encode the parameters of NURBS surfaces. Our evaluation in solid generation and segmentation tasks indicates that the NeuroNURBS performs comparably and, in some cases, superior to UV-grids, but with a significantly improved efficiency: for training the surface autoencoder, GPU consumption is reduced by 86.7%; memory requirement drops by 79.9% for storing 3D solids. Moreover, adapting BrepGen for solid generation with our NeuroNURBS improves the FID from 30.04 to 27.24, and resolves the undulating issue in generated surfaces.