Toward Robust Neural Reconstruction from Sparse Point Sets

Fuente: arXiv
Salvato in:
Dettagli Bibliografici
Autori principali: Ouasfi, Amine, Jena, Shubhendu, Marchand, Eric, Boukhayma, Adnane
Natura: Preprint
Pubblicazione: 2024
Soggetti:
Accesso online:
Tags: Aggiungi Tag
Nessun Tag, puoi essere il primo ad aggiungerne!!
_version_ 1866912165290573824
author Ouasfi, Amine
Jena, Shubhendu
Marchand, Eric
Boukhayma, Adnane
author_facet Ouasfi, Amine
Jena, Shubhendu
Marchand, Eric
Boukhayma, Adnane
contents We consider the challenging problem of learning Signed Distance Functions (SDF) from sparse and noisy 3D point clouds. In contrast to recent methods that depend on smoothness priors, our method, rooted in a distributionally robust optimization (DRO) framework, incorporates a regularization term that leverages samples from the uncertainty regions of the model to improve the learned SDFs. Thanks to tractable dual formulations, we show that this framework enables a stable and efficient optimization of SDFs in the absence of ground truth supervision. Using a variety of synthetic and real data evaluations from different modalities, we show that our DRO based learning framework can improve SDF learning with respect to baselines and the state-of-the-art methods.
format Preprint
id arxiv_https___arxiv_org_abs_2412_16361
institution arXiv
publishDate 2024
record_format arxiv
spellingShingle Toward Robust Neural Reconstruction from Sparse Point Sets
Ouasfi, Amine
Jena, Shubhendu
Marchand, Eric
Boukhayma, Adnane
Computer Vision and Pattern Recognition
We consider the challenging problem of learning Signed Distance Functions (SDF) from sparse and noisy 3D point clouds. In contrast to recent methods that depend on smoothness priors, our method, rooted in a distributionally robust optimization (DRO) framework, incorporates a regularization term that leverages samples from the uncertainty regions of the model to improve the learned SDFs. Thanks to tractable dual formulations, we show that this framework enables a stable and efficient optimization of SDFs in the absence of ground truth supervision. Using a variety of synthetic and real data evaluations from different modalities, we show that our DRO based learning framework can improve SDF learning with respect to baselines and the state-of-the-art methods.
title Toward Robust Neural Reconstruction from Sparse Point Sets
topic Computer Vision and Pattern Recognition
url https://arxiv.org/abs/2412.16361