cs.CVNov 18, 2026

What Do Learned 3D Models Add to Localization from Posed References?

Authors: Jaehyun Kim, Wonseok Kang, Jeongmin Lee, Tae-Wan Kim

Organizations: Seoul National University, Department of Naval Architecture and Ocean Engineering, Seoul, South Korea · DeepFine Co., Ltd., Seoul, South Korea

Abstract

Localizing a query from a few posed reference images avoids building a 3D point map. We ask what the camera predictions of learned 3D models add to geometric estimation from the same references, and where failures arise. Across eight learned models and variants and five test sets, we compare reference budgets, swap pose components, and intervene on the references a model sees. At fine pose-error thresholds, geometric estimation with one additional reference is significantly more accurate than every model on two references on Cambridge and NAVER B1 and never significantly worse, alone or as a fallback behind a shared acceptance gate; with the same three references, the more accurate approach depends on the scene. The gain from the learned translation magnitude depends on reference selection: it improves seven of the eight models on 12-Scenes with the spread-out pairs we select, but none of the five we test with the two top-ranked retrieved references. Where retrieval also returns references from other locations, at NAVER Gangnam Station, models fail at different stages. A robust alignment brings Pi3X and MapAnything within 3.3 pp of adding references as needed. In controlled tests, aligning on the nearby references largely restores VGGT, Pi3X and MapAnything but not Depth Anything 3, and pose-conditioned Pi3X recovers only if the far poses are masked and left out of its input-pose normalization.

Figures & tables

Explore similar work

CardsList
  1. Reference-Induced Consensus for Selective Posed-Reference Visual Localization

    Jul 6, 2026Wonseok Kang, Jaehyun Kim, Jeongmin Lee +1Camera Pose EstimationRobot Localization

  2. TrianguLang: Geometry-Aware Semantic Consensus for Pose-Free 3D Localization

    Mar 9, 2026Bryce Grant, Aryeh Rothenberg, Atri Banerjee +13D Visual GroundingOpen-Vocabulary 3D Segmentation

  3. Learning Proposes, Geometry Disposes: A Modular Framework for Efficient Spatial Reasoning

    Feb 16, 2026Haichao Zhu, Zhaorui Yang, Qian ZhangRelative Pose EstimationDepth Estimation