Gaussian Splatting Feature Fields for (Privacy-Preserving) Visual Localization

Kód výsledku v IS VaVaI

<a href="https://www.isvavai.cz/riv?ss=detail&h=RIV%2F68407700%3A21230%2F25%3A00383813" target="_blank" >RIV/68407700:21230/25:00383813 - isvavai.cz</a>

Nalezeny alternativní kódy

RIV/68407700:21730/25:00383813

Výsledek na webu

<a href="https://doi.org/10.1109/CVPR52734.2025.00109" target="_blank" >https://doi.org/10.1109/CVPR52734.2025.00109</a>

DOI - Digital Object Identifier

<a href="http://dx.doi.org/10.1109/CVPR52734.2025.00109" target="_blank" >10.1109/CVPR52734.2025.00109</a>

Jazyk výsledku

angličtina

Název v původním jazyce

Gaussian Splatting Feature Fields for (Privacy-Preserving) Visual Localization

Popis výsledku v původním jazyce

Visual localization is the task of estimating a camera pose in a known environment. In this paper, we utilize 3D Gaussian Splatting (3DGS)-based representations for accurate and privacy-preserving visual localization. We propose Gaus sian Splatting Feature Fields (GSFFs), a scene represen tation for visual localization that combines an explicit ge ometry model (3DGS) with an implicit feature field. We leverage the dense geometric information and differentiable rasterization algorithm from 3DGS to learn robust feature representations grounded in 3D. In particular, we align a 3D scale-aware feature field and a 2D feature encoder in a common embedding space through a contrastive frame work. Using a 3D structure-informed clustering procedure, wefurther regularize the representation learning and seam lessly convert the features to segmentations, which can be used for privacy-preserving visual localization. Pose refine ment, which involves aligning either feature maps or seg mentations from a query image with those rendered from the GSFFsscenerepresentation, is used to achieve localization. Theresulting privacy- and non-privacy-preserving localiza tion pipelines, evaluated on multiple real-world datasets, show state-of-the-art performances.

Název v anglickém jazyce

Gaussian Splatting Feature Fields for (Privacy-Preserving) Visual Localization

Popis výsledku anglicky

Visual localization is the task of estimating a camera pose in a known environment. In this paper, we utilize 3D Gaussian Splatting (3DGS)-based representations for accurate and privacy-preserving visual localization. We propose Gaus sian Splatting Feature Fields (GSFFs), a scene represen tation for visual localization that combines an explicit ge ometry model (3DGS) with an implicit feature field. We leverage the dense geometric information and differentiable rasterization algorithm from 3DGS to learn robust feature representations grounded in 3D. In particular, we align a 3D scale-aware feature field and a 2D feature encoder in a common embedding space through a contrastive frame work. Using a 3D structure-informed clustering procedure, wefurther regularize the representation learning and seam lessly convert the features to segmentations, which can be used for privacy-preserving visual localization. Pose refine ment, which involves aligning either feature maps or seg mentations from a query image with those rendered from the GSFFsscenerepresentation, is used to achieve localization. Theresulting privacy- and non-privacy-preserving localiza tion pipelines, evaluated on multiple real-world datasets, show state-of-the-art performances.

Druh

D - Stať ve sborníku

CEP obor

—

OECD FORD obor

10201 - Computer sciences, information science, bioinformathics (hardware development to be 2.2, social aspect to be 5.8)

Projekt

—

Návaznosti

S - Specificky vyzkum na vysokych skolach

Rok uplatnění

2025

Kód důvěrnosti údajů

S - Úplné a pravdivé údaje o projektu nepodléhají ochraně podle zvláštních právních předpisů

Název statě ve sborníku

2025 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR)

ISBN

979-8-3315-4364-8

ISSN

1063-6919

e-ISSN

2575-7075

Počet stran výsledku

11

Strana od-do

1082-1092

Název nakladatele

IEEE Computer Society

Místo vydání

Los Alamitos

Místo konání akce

Nashville

Datum konání akce

11. 6. 2025

Typ akce podle státní příslušnosti

WRD - Celosvětová akce

Kód UT WoS článku

001562507801045