Emergent Outlier View Rejection in Visual Geometry Grounded Transformers

Fuente: arXiv
Gespeichert in:
Bibliographische Detailangaben
Hauptverfasser: Han, Jisang, Hong, Sunghwan, Jung, Jaewoo, Jang, Wooseok, An, Honggyu, Wang, Qianqian, Kim, Seungryong, Feng, Chen
Format: Preprint
Veröffentlicht: 2025
Schlagworte:
Online-Zugang:
Tags: Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
_version_ 1866915651839328256
author Han, Jisang
Hong, Sunghwan
Jung, Jaewoo
Jang, Wooseok
An, Honggyu
Wang, Qianqian
Kim, Seungryong
Feng, Chen
author_facet Han, Jisang
Hong, Sunghwan
Jung, Jaewoo
Jang, Wooseok
An, Honggyu
Wang, Qianqian
Kim, Seungryong
Feng, Chen
contents Reliable 3D reconstruction from in-the-wild image collections is often hindered by "noisy" images-irrelevant inputs with little or no view overlap with others. While traditional Structure-from-Motion pipelines handle such cases through geometric verification and outlier rejection, feed-forward 3D reconstruction models lack these explicit mechanisms, leading to degraded performance under in-the-wild conditions. In this paper, we discover that the existing feed-forward reconstruction model, e.g., VGGT, despite lacking explicit outlier-rejection mechanisms or noise-aware training, can inherently distinguish distractor images. Through an in-depth analysis under varying proportions of synthetic distractors, we identify a specific layer that naturally exhibits outlier-suppressing behavior. Further probing reveals that this layer encodes discriminative internal representations that enable an effective noise-filtering capability, which we simply leverage to perform outlier-view rejection in feed-forward 3D reconstruction without any additional fine-tuning or supervision. Extensive experiments on both controlled and in-the-wild datasets demonstrate that this implicit filtering mechanism is consistent and generalizes well across diverse scenarios.
format Preprint
id arxiv_https___arxiv_org_abs_2512_04012
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle Emergent Outlier View Rejection in Visual Geometry Grounded Transformers
Han, Jisang
Hong, Sunghwan
Jung, Jaewoo
Jang, Wooseok
An, Honggyu
Wang, Qianqian
Kim, Seungryong
Feng, Chen
Computer Vision and Pattern Recognition
Reliable 3D reconstruction from in-the-wild image collections is often hindered by "noisy" images-irrelevant inputs with little or no view overlap with others. While traditional Structure-from-Motion pipelines handle such cases through geometric verification and outlier rejection, feed-forward 3D reconstruction models lack these explicit mechanisms, leading to degraded performance under in-the-wild conditions. In this paper, we discover that the existing feed-forward reconstruction model, e.g., VGGT, despite lacking explicit outlier-rejection mechanisms or noise-aware training, can inherently distinguish distractor images. Through an in-depth analysis under varying proportions of synthetic distractors, we identify a specific layer that naturally exhibits outlier-suppressing behavior. Further probing reveals that this layer encodes discriminative internal representations that enable an effective noise-filtering capability, which we simply leverage to perform outlier-view rejection in feed-forward 3D reconstruction without any additional fine-tuning or supervision. Extensive experiments on both controlled and in-the-wild datasets demonstrate that this implicit filtering mechanism is consistent and generalizes well across diverse scenarios.
title Emergent Outlier View Rejection in Visual Geometry Grounded Transformers
topic Computer Vision and Pattern Recognition
url https://arxiv.org/abs/2512.04012