4 Hits in 2.2 sec

UV-Net: Learning from Boundary Representations [article]

Pradeep Kumar Jayaraman, Aditya Sanghi, Joseph G. Lambourne, Karl D.D. Willis, Thomas Davies, Hooman Shayani, Nigel Morris
2021 arXiv   pre-print
We introduce UV-Net, a novel neural network architecture and representation designed to operate directly on Boundary representation (B-rep) data from 3D CAD models.  ...  This leads to a unique and efficient network architecture, UV-Net, that couples image and graph convolutional neural networks in a compute and memory-efficient manner.  ...  Acknowledgments and Disclosure of Funding We thank Karl Willis for helping with processing the ABC dataset and proofreading, and Peter Bentley for helpful suggestions that improved the paper.  ... 
arXiv:2006.10211v2 fatcat:vccqocfodvcihke2kk4lx6mvli

UVStyle-Net: Unsupervised Few-shot Learning of 3D Style Similarity Measure for B-Reps [article]

Peter Meltzer, Hooman Shayani, Amir Khasahmadi, Pradeep Kumar Jayaraman, Aditya Sanghi, Joseph Lambourne
2021 arXiv   pre-print
We propose UVStyle-Net, a style similarity measure for B-Reps that leverages the style signals in the second order statistics of the activations in a pre-trained (unsupervised) 3D encoder, and learns their  ...  We also show it is able to generate meaningful style gradients with respect to the input shape, and that few-shot learning with as few as two positive examples selected by an end-user is sufficient to  ...  They are typically used to describe closed volumes (solids), but can also represent 2D manifolds (sheets) and curve networks (wire bodies).  ... 
arXiv:2105.02961v3 fatcat:sxdznzourzddnbtvaz3g4yy2kq

DeepCAD: A Deep Generative Network for Computer-Aided Design Models [article]

Rundi Wu, Chang Xiao, Changxi Zheng
2021 arXiv   pre-print
To train our network, we create a new CAD dataset consisting of 178,238 models and their CAD construction sequences.  ...  Drawing an analogy between CAD operations and natural language, we propose a CAD generative network based on the Transformer.  ...  This work was partially supported by the National Science Foundation (1910839 and 1816041).  ... 
arXiv:2105.09492v2 fatcat:lpsjxzzdsncmldui57ksxdec5e

AutoMate: A Dataset and Learning Approach for Automatic Mating of CAD Assemblies [article]

Benjamin Jones, Dalton Hildreth, Duowen Chen, Ilya Baran, Vladimir G. Kim, Adriana Schulz
2021 arXiv   pre-print
We propose SB-GCN, a representation learning scheme on BREPs that retains the topological structure of parts, and use these learned representations to predict CAD type mates.  ...  While point density around sharp features presents a challenge for PointNet features, other models such as MeshCNN [Hanocka et al. 2019] or the curve and surface convolution features of UV-Net [Jayaraman  ...  To identify identical parts and assemblies, we created a BREP fingerprint from the number of each type of topological entity and the moment of inertia tensor and center of mass of the represented solid  ... 
arXiv:2105.12238v2 fatcat:5gr7oorgqvcuhfigzct6nuipr4