Neural Splines: Fitting 3D Surfaces with Infinitely-Wide Neural Networks
Preprint
Abstract
We present Neural Splines, a technique for 3D surface reconstruction that is based on random feature kernels arising from infinitely-wide shallow ReLU networks. Our method achieves state-of-the-art results, outperforming recent neural network-based techniques and widely used Poisson Surface Reconstruction (which, as we demonstrate, can also be viewed as a type of kernel method). Because our approach is based on a simple kernel formulation, it is easy to analyze and can be accelerated by general techniques designed for kernel-based learning. We provide explicit analytical expressions for our kernel and argue that our formulation can be seen as a generalization of cubic spline interpolation to higher dimensions. In particular, the RKHS norm associated with Neural Splines biases toward smooth interpolants.Keywords
All Related Versions
This publication has 13 references indexed in Scilit:
- Deep Manifold Prior2021
- Deep Local Shapes: Learning Local SDF Priors for Detailed 3D Reconstruction2020
- CvxNet: Learnable Convex Decomposition2020
- Local Deep Implicit Functions for 3D Shape2020
- SAL: Sign Agnostic Learning of Shapes From Raw Data2020
- BSP-Net: Generating Compact Meshes via Binary Space Partitioning2020
- Learning Shape Templates With Structured Implicit Functions2019
- Occupancy Networks: Learning 3D Reconstruction in Function Space2019
- Sparse surface reconstruction with adaptive partition of unity and radial basis functionsGraphical Models, 2006
- A fast algorithm for particle simulationsJournal of Computational Physics, 1987