A Finite Difference Approximation of Second Order Regularization of Neural-SDFs

Fuente: arXiv
Gespeichert in:
Bibliographische Detailangaben
Hauptverfasser: Yin, Haotian, Plocharski, Aleksander, Wlodarczyk, Michal Jan, Musialski, Przemyslaw
Format: Preprint
Veröffentlicht: 2025
Schlagworte:
Online-Zugang:
Tags: Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
_version_ 1866918198363815936
author Yin, Haotian
Plocharski, Aleksander
Wlodarczyk, Michal Jan
Musialski, Przemyslaw
author_facet Yin, Haotian
Plocharski, Aleksander
Wlodarczyk, Michal Jan
Musialski, Przemyslaw
contents We introduce a finite-difference framework for curvature regularization in neural signed distance field (SDF) learning. Existing approaches enforce curvature priors using full Hessian information obtained via second-order automatic differentiation, which is accurate but computationally expensive. Others reduced this overhead by avoiding explicit Hessian assembly, but still required higher-order differentiation. In contrast, our method replaces these operations with lightweight finite-difference stencils that approximate second derivatives using the well known Taylor expansion with a truncation error of O(h^2), and can serve as drop-in replacements for Gaussian curvature and rank-deficiency losses. Experiments demonstrate that our finite-difference variants achieve reconstruction fidelity comparable to their automatic-differentiation counterparts, while reducing GPU memory usage and training time by up to a factor of two. Additional tests on sparse, incomplete, and non-CAD data confirm that the proposed formulation is robust and general, offering an efficient and scalable alternative for curvature-aware SDF learning.
format Preprint
id arxiv_https___arxiv_org_abs_2511_08980
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle A Finite Difference Approximation of Second Order Regularization of Neural-SDFs
Yin, Haotian
Plocharski, Aleksander
Wlodarczyk, Michal Jan
Musialski, Przemyslaw
Graphics
Computer Vision and Pattern Recognition
Machine Learning
We introduce a finite-difference framework for curvature regularization in neural signed distance field (SDF) learning. Existing approaches enforce curvature priors using full Hessian information obtained via second-order automatic differentiation, which is accurate but computationally expensive. Others reduced this overhead by avoiding explicit Hessian assembly, but still required higher-order differentiation. In contrast, our method replaces these operations with lightweight finite-difference stencils that approximate second derivatives using the well known Taylor expansion with a truncation error of O(h^2), and can serve as drop-in replacements for Gaussian curvature and rank-deficiency losses. Experiments demonstrate that our finite-difference variants achieve reconstruction fidelity comparable to their automatic-differentiation counterparts, while reducing GPU memory usage and training time by up to a factor of two. Additional tests on sparse, incomplete, and non-CAD data confirm that the proposed formulation is robust and general, offering an efficient and scalable alternative for curvature-aware SDF learning.
title A Finite Difference Approximation of Second Order Regularization of Neural-SDFs
topic Graphics
Computer Vision and Pattern Recognition
Machine Learning
url https://arxiv.org/abs/2511.08980