Corpus ID: 220041556

Neural Splines: Fitting 3D Surfaces with Infinitely-Wide Neural Networks

@inproceedings{Williams2021NeuralSF,
  title={Neural Splines: Fitting 3D Surfaces with Infinitely-Wide Neural Networks},
  author={Francis Williams and Matthew Trager and Joan Bruna and D. Zorin},
  booktitle={CVPR},
  year={2021}
}
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 Screened Poisson Surface Reconstruction and modern neural network based techniques. Because our approach is based on a simple kernel formulation, it is fast to run and easy to analyze. We provide explicit analytical expressions for our kernel and argue that our formulation can… Expand
Phase Transitions, Distance Functions, and Implicit Neural Representations
TLDR
Inspiration from the theory of phase transitions of fluids is drawn and a loss for training INRs is suggested that learns a density function that converges to a proper occupancy function, while its log transform converging to a distance function. Expand
Events
The inaugural installment of the congress brings together the expertise of eight leading engineering professional societies and the National Academies to offer a dynamic, cross-sector learning andExpand

References

SHOWING 1-10 OF 47 REFERENCES
Reconstruction and representation of 3D objects with radial basis functions
TLDR
It is shown that the RBF representation has advantages for mesh simplification and remeshing applications, and a greedy algorithm in the fitting process reduces the number of RBF centers required to represent a surface and results in significant compression and further computational advantages. Expand
Screened poisson surface reconstruction
TLDR
This work extends Poisson surface reconstruction to explicitly incorporate the points as interpolation constraints and presents several algorithmic improvements that together reduce the time complexity of the solver to linear in the number of points, thereby enabling faster, higher-quality surface reconstructions. Expand
Deep Geometric Prior for Surface Reconstruction
TLDR
This work proposes the use of a deep neural network as a geometric prior for surface reconstruction, and overfit a neural network representing a local chart parameterization to part of an input point cloud using the Wasserstein distance as a measure of approximation. Expand
Deep Manifold Prior
We present a prior for manifold structured data, such as surfaces of 3D shapes, where deep neural networks are adopted to reconstruct a target shape using gradient descent starting from a randomExpand
Sparse surface reconstruction with adaptive partition of unity and radial basis functions
A new implicit surface fitting method for surface reconstruction from scattered point data is proposed. The method combines an adaptive partition of unity approximation with least-squares RBF fittingExpand
Point2Mesh: A Self-Prior for Deformable Meshes
TLDR
This paper introduces Point2Mesh, a technique for reconstructing a surface mesh from an input point cloud that is robust to non-ideal conditions, and shows that shrink-wrapping a point cloud with a self-prior converges to a desirable solution. Expand
Occupancy Networks: Learning 3D Reconstruction in Function Space
TLDR
This paper proposes Occupancy Networks, a new representation for learning-based 3D reconstruction methods that encodes a description of the 3D output at infinite resolution without excessive memory footprint, and validate that the representation can efficiently encode 3D structure and can be inferred from various kinds of input. Expand
Surface reconstruction from unorganized points
TLDR
A general method for automatic reconstruction of accurate, concise, piecewise smooth surfaces from unorganized 3D points that is able to automatically infer the topological type of the surface, its geometry, and the presence and location of features such as boundaries, creases, and corners. Expand
DeepSDF: Learning Continuous Signed Distance Functions for Shape Representation
TLDR
This work introduces DeepSDF, a learned continuous Signed Distance Function (SDF) representation of a class of shapes that enables high quality shape representation, interpolation and completion from partial and noisy 3D input data. Expand
Algebraic point set surfaces
TLDR
This paper presents a new Point Set Surface (PSS) definition based on moving least squares (MLS) fitting of algebraic spheres, and presents an novel normal estimation procedure which exploits the properties of the spherical fit for both direction estimation and orientation propagation. Expand
...
1
2
3
4
5
...