GenRecon: Bridging Generative Priors for Multi-View 3D Scene Reconstruction

Fuente: arXiv
Salvato in:
Dettagli Bibliografici
Autori principali: Schmid, Katharina, von Lützow, Nicolas, Hladký, Jozef, Dai, Angela, Nießner, Matthias
Natura: Preprint
Pubblicazione: 2026
Soggetti:
Accesso online:
Tags: Aggiungi Tag
Nessun Tag, puoi essere il primo ad aggiungerne!!
_version_ 1866913157229838336
author Schmid, Katharina
von Lützow, Nicolas
Hladký, Jozef
Dai, Angela
Nießner, Matthias
author_facet Schmid, Katharina
von Lützow, Nicolas
Hladký, Jozef
Dai, Angela
Nießner, Matthias
contents We introduce a new approach to high-fidelity 3D scene reconstruction from multi-view RGB images that tightly couples reconstruction with a strong generative 3D prior. We cast scene reconstruction as conditional 3D generation over a set of spatially-localized, overlapping chunks that together tile the scene, scaling generation to large scene extents. Crucially, we inherit the fidelity and completeness of state-of-the-art generative shape models -- we use Trellis.2 as an example -- which we generalize to the scene level. To this end, we propose a projection-based conditioning mechanism that lifts posed multi-view image features into a coherent 3D representation aligned with the generative model, independent of view ordering and spatially anchored to the scene, yielding high-fidelity, multi-view consistent generated geometry. This enables lifting the strong object-level prior of Trellis.2 to multi-view, scene-scale generation, producing faithful, editable PBR mesh reconstructions of indoor environments. As a result, we obtain high-fidelity results that outperform cutting-edge reconstruction methods by 16%.
format Preprint
id arxiv_https___arxiv_org_abs_2605_23888
institution arXiv
publishDate 2026
record_format arxiv
spellingShingle GenRecon: Bridging Generative Priors for Multi-View 3D Scene Reconstruction
Schmid, Katharina
von Lützow, Nicolas
Hladký, Jozef
Dai, Angela
Nießner, Matthias
Computer Vision and Pattern Recognition
We introduce a new approach to high-fidelity 3D scene reconstruction from multi-view RGB images that tightly couples reconstruction with a strong generative 3D prior. We cast scene reconstruction as conditional 3D generation over a set of spatially-localized, overlapping chunks that together tile the scene, scaling generation to large scene extents. Crucially, we inherit the fidelity and completeness of state-of-the-art generative shape models -- we use Trellis.2 as an example -- which we generalize to the scene level. To this end, we propose a projection-based conditioning mechanism that lifts posed multi-view image features into a coherent 3D representation aligned with the generative model, independent of view ordering and spatially anchored to the scene, yielding high-fidelity, multi-view consistent generated geometry. This enables lifting the strong object-level prior of Trellis.2 to multi-view, scene-scale generation, producing faithful, editable PBR mesh reconstructions of indoor environments. As a result, we obtain high-fidelity results that outperform cutting-edge reconstruction methods by 16%.
title GenRecon: Bridging Generative Priors for Multi-View 3D Scene Reconstruction
topic Computer Vision and Pattern Recognition
url https://arxiv.org/abs/2605.23888