Abstract
Boundary representations (B-reps) using Non-Uniform Rational B-splines(NURBS) are the de facto standard used in CAD, but their utility in deeplearning-based approaches is not well researched. We propose a differentiableNURBS module to integrate the NURBS representation of CAD models with deeplearning methods. We mathematically define the derivatives of the NURBS curvesor surfaces with respect to the input parameters. These derivatives are used todefine an approximate Jacobian that can be used to perform the "backward"evaluation used while training deep learning models. We have implemented ourNURBS module using GPU-accelerated algorithms and integrated it with PyTorch, apopular deep learning framework. We demonstrate the efficacy of our NURBSmodule in performing CAD operations such as curve or surface fitting andsurface offsetting. Further, we show its utility in deep learning forunsupervised point cloud reconstruction. These examples show that our moduleperforms better for certain deep learning frameworks and can be directlyintegrated with any deep-learning framework requiring NURBS.