MVSAnywhere: Zero-Shot Multi-View Stereo

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Izquierdo, Sergio, Sayed, Mohamed, Firman, Michael, Garcia-Hernando, Guillermo, Turmukhambetov, Daniyar, Civera, Javier, Mac Aodha, Oisin, Brostow, Gabriel, Watson, Jamie
Format: Preprint
Published: 2025
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866909556087455744
author Izquierdo, Sergio
Sayed, Mohamed
Firman, Michael
Garcia-Hernando, Guillermo
Turmukhambetov, Daniyar
Civera, Javier
Mac Aodha, Oisin
Brostow, Gabriel
Watson, Jamie
author_facet Izquierdo, Sergio
Sayed, Mohamed
Firman, Michael
Garcia-Hernando, Guillermo
Turmukhambetov, Daniyar
Civera, Javier
Mac Aodha, Oisin
Brostow, Gabriel
Watson, Jamie
contents Computing accurate depth from multiple views is a fundamental and longstanding challenge in computer vision. However, most existing approaches do not generalize well across different domains and scene types (e.g. indoor vs. outdoor). Training a general-purpose multi-view stereo model is challenging and raises several questions, e.g. how to best make use of transformer-based architectures, how to incorporate additional metadata when there is a variable number of input views, and how to estimate the range of valid depths which can vary considerably across different scenes and is typically not known a priori? To address these issues, we introduce MVSA, a novel and versatile Multi-View Stereo architecture that aims to work Anywhere by generalizing across diverse domains and depth ranges. MVSA combines monocular and multi-view cues with an adaptive cost volume to deal with scale-related issues. We demonstrate state-of-the-art zero-shot depth estimation on the Robust Multi-View Depth Benchmark, surpassing existing multi-view stereo and monocular baselines.
format Preprint
id arxiv_https___arxiv_org_abs_2503_22430
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle MVSAnywhere: Zero-Shot Multi-View Stereo
Izquierdo, Sergio
Sayed, Mohamed
Firman, Michael
Garcia-Hernando, Guillermo
Turmukhambetov, Daniyar
Civera, Javier
Mac Aodha, Oisin
Brostow, Gabriel
Watson, Jamie
Computer Vision and Pattern Recognition
Computing accurate depth from multiple views is a fundamental and longstanding challenge in computer vision. However, most existing approaches do not generalize well across different domains and scene types (e.g. indoor vs. outdoor). Training a general-purpose multi-view stereo model is challenging and raises several questions, e.g. how to best make use of transformer-based architectures, how to incorporate additional metadata when there is a variable number of input views, and how to estimate the range of valid depths which can vary considerably across different scenes and is typically not known a priori? To address these issues, we introduce MVSA, a novel and versatile Multi-View Stereo architecture that aims to work Anywhere by generalizing across diverse domains and depth ranges. MVSA combines monocular and multi-view cues with an adaptive cost volume to deal with scale-related issues. We demonstrate state-of-the-art zero-shot depth estimation on the Robust Multi-View Depth Benchmark, surpassing existing multi-view stereo and monocular baselines.
title MVSAnywhere: Zero-Shot Multi-View Stereo
topic Computer Vision and Pattern Recognition
url https://arxiv.org/abs/2503.22430