Deep Multi-View Stereo Gone Wild - Archive ouverte HAL Accéder directement au contenu
Communication Dans Un Congrès Année : 2022

Deep Multi-View Stereo Gone Wild

Résumé

Deep multi-view stereo (MVS) methods have been developed and extensively compared on simple datasets, where they now outperform classical approaches. In this paper, we ask whether the conclusions reached in controlled scenarios are still valid when working with Internet photo collections. We propose a methodology for evaluation and explore the influence of three aspects of deep MVS methods: network architecture, training data, and supervision. We make several key observations, which we extensively validate quantitatively and qualitatively, both for depth prediction and complete 3D reconstructions. First, complex unsupervised approaches cannot train on data in the wild. Our new approach makes it possible with three key elements: upsampling the output, softmin based aggregation and a single reconstruction loss. Second, supervised deep depthmap-based MVS methods are state-of-the art for reconstruction of few internet images. Finally, our evaluation provides very different results than usual ones. This shows that evaluation in uncontrolled scenarios is important for new architectures.
Fichier principal
Vignette du fichier
Francois2020.pdf (8.92 Mo) Télécharger le fichier
Origine : Fichiers produits par l'(les) auteur(s)

Dates et versions

hal-04315771 , version 1 (30-11-2023)

Identifiants

Citer

Francois Darmon, Benedicte Bascle, Jean-Clement Devaux, Pascal Monasse, Mathieu Aubry. Deep Multi-View Stereo Gone Wild. 2021 International Conference on 3D Vision (3DV), Dec 2021, London, United Kingdom. pp.484-493, ⟨10.1109/3DV53792.2021.00058⟩. ⟨hal-04315771⟩
8 Consultations
3 Téléchargements

Altmetric

Partager

Gmail Facebook X LinkedIn More