pith. sign in

arxiv: 2605.04730 · v1 · submitted 2026-05-06 · 💻 cs.CV

ULF-Loc: Unbiased Landmark Feature for Robust Visual Localization with 3D Gaussian Splatting

Yingdong Gu , Shaocheng Yan , Zhenjun Zhao , Yuan Kou , Jianxin Luo , Pengcheng Shi , Jiayuan Li This is my paper

Pith reviewed 2026-05-08 17:22 UTC · model grok-4.3

classification 💻 cs.CV
keywords visual localization3D Gaussian splattingfeature biaslandmark matchingcamera pose estimationaugmented realityautonomous navigation
0
0 comments X

The pith

Alpha-blending in 3D Gaussian Splatting creates biased features that cause mismatches in visual localization, which ULF-Loc corrects via geometry-weighted fusion and consensus sampling.

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The paper establishes that standard optimization of features inside 3D Gaussian Splatting entangles each Gaussian with its neighbors through alpha-blending, producing point features that are systematically unsuitable for direct 2D-to-3D matching. This entanglement bias is shown to be the source of frequent mismatches during localization. ULF-Loc replaces the biased optimization with geometry-weighted feature fusion, selects reliable landmarks via keypoint-consensus sampling, and applies local geometric consistency checks to reject remaining artifacts. The resulting unbiased features deliver higher accuracy on standard benchmarks while using substantially less training time and memory.

Core claim

The central claim is that alpha-blending optimization in 3DGS inherently introduces bias into learned 3D point features through entanglement with neighboring Gaussians, rendering them unsuitable for precise matching. Replacing this process with geometry-weighted feature fusion produces unbiased landmark features, and supplementing it with keypoint-consensus sampling plus local geometric consistency verification yields robust visual localization.

What carries the argument

Geometry-weighted feature fusion, which decouples each Gaussian's feature from its neighbors by reweighting contributions according to geometric properties instead of alpha-blending weights.

If this is right

  • On the Cambridge Landmarks dataset the approach reduces mean median translation error by 17 percent relative to prior state-of-the-art methods.
  • Training requires only one-tenth the time and one-sixth the GPU memory of the previous leading method.
  • The same pipeline supports reliable camera pose estimation for augmented reality and autonomous navigation tasks.
  • Local geometric consistency verification removes mismatches that arise from rendering artifacts after fusion.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

  • The same bias-correction principle may apply to feature learning inside other differentiable rendering pipelines beyond 3D Gaussian Splatting.
  • Efficiency gains could enable on-device visual localization on hardware with limited compute and memory.
  • Evaluating the method on datasets containing strong lighting changes or moving objects would test whether the unbiased features remain stable outside controlled landmark scenes.

Load-bearing premise

The assumption that alpha-blending entanglement is the dominant source of feature mismatches and that geometry-weighted fusion plus consensus sampling removes this bias without discarding useful information or introducing new artifacts.

What would settle it

If a controlled experiment learns Gaussian features without alpha-blending (for example by direct per-Gaussian supervision) and the proposed fusion method then shows no accuracy gain over simple matching, the claim that entanglement is the primary bias source would be falsified.

Figures

Figures reproduced from arXiv: 2605.04730 by Jianxin Luo, Jiayuan Li, Pengcheng Shi, Shaocheng Yan, Yingdong Gu, Yuan Kou, Zhenjun Zhao.

Figure 1
Figure 1. Figure 1: Overview of ULF-Loc. ULF-Loc first constructs a set of landmarks G˜ with unbiased features via K.C. Sampling and GWFF. The camera pose is estimated in a coarse-to-fine manner: it computes an initial pose Pcoarse by matching landmarks to 2D keypoints, then refines it to Pf ine through dense feature matching. During refinement, the LGCV rejects mismatches caused by rendering artifacts. 3. Method We present a… view at source ↗
Figure 2
Figure 2. Figure 2: Feature Distance Distributions on 7Scenes Chess. Our feature fusion yields a sharply peaked distribution at small distances, indicating superior 2D-3D consistency, while Feature￾3DGS shows a broad distribution, revealing its inherent bias. where fi is the 3D feature of gi (from Feature-3DGS [70] or our feature fusion) view at source ↗
Figure 3
Figure 3. Figure 3: Geometry-Weighted Feature Fusion. For each land￾mark Gaussian (center \mu _i ), we compute fusion weights based on the cosine similarity between its normal \protect \bm {n}_i and viewing directions di,k from multiple cameras. Query Image Render Image ℳ view at source ↗
Figure 4
Figure 4. Figure 4: Local Geometric Consistency Verification. Each match pair (xi, yi) forms triangular pairs with its k-NN. Each tri￾angular pair is then checked for two geometric constraints. Gaussian scene \protect \mathcal {G}. We then extract feature maps from both the rendered image \protect \hat {I}_s and the query image I_q using \protect \mathcal {F} for dense feature matching. Inspired by [18, 54, 63], we adopt a co… view at source ↗
Figure 5
Figure 5. Figure 5: Recall Rate for Different Numbers of Landmarks view at source ↗
Figure 6
Figure 6. Figure 6: Qualitative Comparison of LGCV. LGCV enhances the accuracy of pose estimation by eliminating mismatches. achieves better localization results by obtaining unbiased and accurate 3D features. Further comparison between (#7 and #8) shows that the geometry-weighted mechanism ef￾fectively improves performance, as it fully considers the im￾pact of viewpoint variation on 3D feature representation. Efficacy of Loc… view at source ↗
read the original abstract

Visual localization is a core technology for augmented reality and autonomous navigation. Recent methods combine the efficient rendering of 3D Gaussian Splatting (3DGS) with feature-based localization. These methods rely on direct matching between 2D query features and the 3D Gaussian feature field, but this often results in mismatches due to an inherent bias in the learned Gaussian feature. We theoretically analyze the feature learning process in 3DGS, revealing that the widely adopted $\alpha$-blending optimization inherently introduces bias into 3D point features. This bias stems from the entanglement between individual Gaussians and their neighboring Gaussians, making the learned features unsuitable for precise matching tasks. Motivated by these findings, we propose ULF-Loc, an unbiased landmark feature framework that replaces biased feature optimization with geometry-weighted feature fusion. We further introduce keypoint-consensus landmark sampling to select reliable Gaussians and local geometric consistency verification to reject mismatches caused by rendering artifacts. On the Cambridge Landmarks dataset, ULF-Loc reduces the mean median translation error by 17\% compared to the state-of-the-art, while achieving superior efficiency with only 1/10 the training time and 1/6 the GPU memory of STDLoc.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Referee Report

3 major / 1 minor

Summary. The paper claims that α-blending optimization in 3D Gaussian Splatting inherently biases learned point features through entanglement with neighboring Gaussians, rendering them unsuitable for precise 2D-3D matching in visual localization. To correct this, ULF-Loc replaces direct feature optimization with geometry-weighted feature fusion, adds keypoint-consensus landmark sampling to select reliable Gaussians, and applies local geometric consistency verification to reject rendering artifacts. On the Cambridge Landmarks dataset the method reports a 17% reduction in mean median translation error versus prior state-of-the-art while using 1/10 the training time and 1/6 the GPU memory of STDLoc.

Significance. If the bias analysis is correct and the observed gains are shown to arise specifically from removal of α-blending entanglement rather than from the auxiliary sampling or verification stages, the work would supply a principled, efficient route to more reliable feature fields for 3DGS-based localization. The efficiency numbers, if reproducible, would also be practically valuable for AR and navigation pipelines that must train or update maps on modest hardware.

major comments (3)
  1. [§3] §3 (theoretical analysis): the claim that α-blending optimization is the dominant source of feature bias is asserted but the full derivation isolating the entanglement term from other optimization effects is not provided; without it the central premise that geometry-weighted fusion removes the primary mismatch cause cannot be verified.
  2. [§5] §5 / Table 1 (Cambridge Landmarks results): the 17% mean-median translation error reduction is reported without component-wise ablations (geometry-weighted fusion alone, consensus sampling alone, verification alone) or inlier-ratio measurements under controlled rendering; therefore it is impossible to attribute the gain to bias removal rather than to the sampling or verification modules.
  3. [§5] §5: no error bars, standard deviations, or statistical significance tests accompany the quantitative localization numbers, weakening confidence that the reported improvement is robust across random seeds or scene variations.
minor comments (1)
  1. [Abstract] Abstract: the phrase 'mean median translation error' is ambiguous; clarify whether it denotes the average of per-scene median errors or another aggregate.

Simulated Author's Rebuttal

3 responses · 0 unresolved

We thank the referee for the constructive and detailed feedback. The comments highlight important areas for strengthening the theoretical grounding and empirical validation of our claims. We address each major comment below and will revise the manuscript accordingly.

read point-by-point responses

  1. Referee: [§3] §3 (theoretical analysis): the claim that α-blending optimization is the dominant source of feature bias is asserted but the full derivation isolating the entanglement term from other optimization effects is not provided; without it the central premise that geometry-weighted fusion removes the primary mismatch cause cannot be verified.

    Authors: We agree that a more explicit isolation of the entanglement term would make the analysis more rigorous. Section 3 derives the bias from the α-blending formulation and shows how neighboring Gaussians affect the optimized features, but we will expand this section in the revision to include the complete step-by-step derivation that separates the entanglement contribution from other optimization dynamics. This will directly support the premise that geometry-weighted fusion targets the dominant bias source. revision: yes

  2. Referee: [§5] §5 / Table 1 (Cambridge Landmarks results): the 17% mean-median translation error reduction is reported without component-wise ablations (geometry-weighted fusion alone, consensus sampling alone, verification alone) or inlier-ratio measurements under controlled rendering; therefore it is impossible to attribute the gain to bias removal rather than to the sampling or verification modules.

    Authors: We recognize that component-wise ablations are necessary to attribute gains specifically to bias removal. The current results emphasize the full pipeline, but we will add these ablations in the revised manuscript, reporting performance for geometry-weighted fusion alone, consensus sampling alone, and verification alone, along with inlier-ratio measurements under controlled rendering to isolate the contribution of each module. revision: yes

  3. Referee: [§5] §5: no error bars, standard deviations, or statistical significance tests accompany the quantitative localization numbers, weakening confidence that the reported improvement is robust across random seeds or scene variations.

    Authors: We concur that variability measures would increase confidence in the results. In the revised version we will augment Table 1 and the §5 results with error bars, standard deviations computed over multiple random seeds, and statistical significance tests to demonstrate that the reported improvements hold robustly across seeds and scene variations. revision: yes

Circularity Check

0 steps flagged

No circularity: theoretical analysis and design choices remain independent of fitted outputs

full rationale

The paper's chain begins with a claimed theoretical analysis of α-blending bias in 3DGS feature learning, followed by independent proposals for geometry-weighted fusion, consensus sampling, and consistency verification. None of these steps reduce by construction to the input data or to quantities fitted from the same data; the methods are presented as motivated design choices rather than reparameterizations of the bias term. No self-citation load-bearing, self-definitional loops, or fitted-input-renamed-as-prediction patterns appear in the provided text. The 17% error reduction is reported as an empirical outcome of the full pipeline, not forced by the analysis itself. This is the common honest case of a self-contained proposal.

Axiom & Free-Parameter Ledger

0 free parameters · 0 axioms · 0 invented entities

Abstract provides no explicit free parameters, axioms, or invented entities; the framework rests on the unstated premise that geometry weighting can isolate unbiased features.

pith-pipeline@v0.9.0 · 5540 in / 1149 out tokens · 23718 ms · 2026-05-08T17:22:01.156809+00:00 · methodology

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Reference graph

Works this paper leans on

70 extracted references · 70 canonical work pages

  1. [1]

    Map-free visual relocalization: Metric pose relative to a single image

    Eduardo Arnold, Jamie Wynn, Sara Vicente, Guillermo Garcia-Hernando, Aron Monszpart, Victor Prisacariu, Dani- yar Turmukhambetov, and Eric Brachmann. Map-free visual relocalization: Metric pose relative to a single image. In European Conference on Computer Vision, pages 690–708. Springer, 2022. 2

    work page 2022

  2. [2]

    Gsloc: Visual localization with 3d gaussian splatting

    Kazii Botashev, Vladislav Pyatov, Gonzalo Ferrer, and Sta- matios Lefkimmiatis. Gsloc: Visual localization with 3d gaussian splatting. In2024 IEEE/RSJ International Confer- ence on Intelligent Robots and Systems (IROS), pages 5664–

    work page

  3. [3]

    Visual camera re- localization from rgb and rgb-d images using dsac.IEEE transactions on pattern analysis and machine intelligence, 44(9):5847–5865, 2021

    Eric Brachmann and Carsten Rother. Visual camera re- localization from rgb and rgb-d images using dsac.IEEE transactions on pattern analysis and machine intelligence, 44(9):5847–5865, 2021. 2, 6, 7, 12

    work page 2021

  4. [4]

    On the limits of pseudo ground truth in vi- sual camera re-localisation

    Eric Brachmann, Martin Humenberger, Carsten Rother, and Torsten Sattler. On the limits of pseudo ground truth in vi- sual camera re-localisation. InProceedings of the IEEE/CVF International Conference on Computer Vision, pages 6218– 6228, 2021. 6

    work page 2021

  5. [5]

    Accelerated coordinate encoding: Learning to relocalize in minutes using rgb and poses

    Eric Brachmann, Tommaso Cavallari, and Victor Adrian Prisacariu. Accelerated coordinate encoding: Learning to relocalize in minutes using rgb and poses. InProceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pages 5044–5053, 2023. 1, 2, 6, 7, 12

    work page 2023

  6. [6]

    Ace-g: Improving generalization of scene coor- dinate regression through query pre-training.arXiv preprint arXiv:2510.11605, 2025

    Leonard Bruns, Axel Barroso-Laguna, Tommaso Cavallari, Sowmya Munukutla, Victor Adrian Prisacariu, Eric Brach- mann, et al. Ace-g: Improving generalization of scene coor- dinate regression through query pre-training.arXiv preprint arXiv:2510.11605, 2025. 2

    work page arXiv 2025

  7. [7]

    Quantifying and alleviating co-adaptation in sparse-view 3d gaussian splatting,

    Kangjie Chen, Yingji Zhong, Zhihao Li, Jiaqi Lin, Youyu Chen, Minghan Qin, and Haoqian Wang. Quantifying and alleviating co-adaptation in sparse-view 3d gaussian splat- ting.arXiv preprint arXiv:2508.12720, 2025. 5

    work page arXiv 2025

  8. [8]

    Leveraging neural radiance fields for uncertainty-aware vi- sual localization

    Le Chen, Weirong Chen, Rui Wang, and Marc Pollefeys. Leveraging neural radiance fields for uncertainty-aware vi- sual localization. In2024 IEEE International Conference on Robotics and Automation (ICRA), pages 6298–6305. IEEE,

    work page

  9. [9]

    Dfnet: Enhance absolute pose regression with direct feature matching

    Shuai Chen, Xinghui Li, Zirui Wang, and Victor A Prisacariu. Dfnet: Enhance absolute pose regression with direct feature matching. InEuropean Conference on Com- puter Vision, pages 1–17. Springer, 2022. 2, 6, 7

    work page 2022

  10. [10]

    Neural refine- ment for absolute pose regression with feature synthesis

    Shuai Chen, Yash Bhalgat, Xinghui Li, Jia-Wang Bian, Kejie Li, Zirui Wang, and Victor Adrian Prisacariu. Neural refine- ment for absolute pose regression with feature synthesis. In Proceedings of the IEEE/CVF Conference on Computer Vi- sion and Pattern Recognition, pages 20987–20996, 2024. 1, 2, 6, 7

    work page 2024

  11. [11]

    Map-relative pose regression for visual re-localization

    Shuai Chen, Tommaso Cavallari, Victor Adrian Prisacariu, and Eric Brachmann. Map-relative pose regression for visual re-localization. InProceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pages 20665– 20674, 2024. 1, 2, 6, 7, 12

    work page 2024

  12. [12]

    Masked-attention mask transformer for universal image segmentation

    Bowen Cheng, Ishan Misra, Alexander G Schwing, Alexan- der Kirillov, and Rohit Girdhar. Masked-attention mask transformer for universal image segmentation. InProceed- ings of the IEEE/CVF conference on computer vision and pattern recognition, pages 1290–1299, 2022. 11

    work page 2022

  13. [13]

    Superpoint: Self-supervised interest point detection and description

    Daniel DeTone, Tomasz Malisiewicz, and Andrew Rabi- novich. Superpoint: Self-supervised interest point detection and description. InProceedings of the IEEE conference on computer vision and pattern recognition workshops, pages 224–236, 2018. 2, 6, 8

    work page 2018

  14. [14]

    Revisiting the p3p problem

    Yaqing Ding, Jian Yang, Viktor Larsson, Carl Olsson, and Kalle ˚Astr¨om. Revisiting the p3p problem. InProceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pages 4872–4880, 2023. 1, 6

    work page 2023

  15. [15]

    Reloc3r: Large-scale training of relative camera pose regression for generalizable, fast, and accurate visual localization

    Siyan Dong, Shuzhe Wang, Shaohui Liu, Lulu Cai, Qingnan Fan, Juho Kannala, and Yanchao Yang. Reloc3r: Large-scale training of relative camera pose regression for generalizable, fast, and accurate visual localization. InProceedings of the Computer Vision and Pattern Recognition Conference, pages 16739–16752, 2025. 1

    work page 2025

  16. [16]

    D2- net: A trainable cnn for joint description and detection of local features

    Mihai Dusmanu, Ignacio Rocco, Tomas Pajdla, Marc Polle- feys, Josef Sivic, Akihiko Torii, and Torsten Sattler. D2- net: A trainable cnn for joint description and detection of local features. InProceedings of the ieee/cvf conference on computer vision and pattern recognition, pages 8092–8101,

    work page

  17. [17]

    Random sample consensus: a paradigm for model fitting with applications to image analysis and automated cartography.Communications of the ACM, 24(6):381–395, 1981

    Martin A Fischler and Robert C Bolles. Random sample consensus: a paradigm for model fitting with applications to image analysis and automated cartography.Communications of the ACM, 24(6):381–395, 1981. 1, 2, 5, 6

    work page 1981

  18. [18]

    From sparse to dense: Camera relocaliza- tion with scene-specific detector from feature gaussian splat- ting

    Zhiwei Huang, Hailin Yu, Yichun Shentu, Jin Yuan, and Guofeng Zhang. From sparse to dense: Camera relocaliza- tion with scene-specific detector from feature gaussian splat- ting. InProceedings of the Computer Vision and Pattern Recognition Conference, pages 27059–27069, 2025. 1, 2, 3, 5, 6, 7, 11, 12

    work page 2025

  19. [19]

    Robust image matching via local graph structure consensus

    Xingyu Jiang, Yifan Xia, Xiao-Ping Zhang, and Jiayi Ma. Robust image matching via local graph structure consensus. Pattern Recognition, 126:108588, 2022. 5

    work page 2022

  20. [20]

    Posenet: A convolutional network for real-time 6-dof cam- era relocalization

    Alex Kendall, Matthew Grimes, and Roberto Cipolla. Posenet: A convolutional network for real-time 6-dof cam- era relocalization. InProceedings of the IEEE international conference on computer vision, pages 2938–2946, 2015. 1, 2, 6

    work page 2015

  21. [21]

    3d gaussian splatting for real-time radiance field rendering.ACM Trans

    Bernhard Kerbl, Georgios Kopanas, Thomas Leimk ¨uhler, and George Drettakis. 3d gaussian splatting for real-time radiance field rendering.ACM Trans. Graph., 42(4):139–1,

    work page

  22. [22]

    Segment any- thing

    Alexander Kirillov, Eric Mintun, Nikhila Ravi, Hanzi Mao, Chloe Rolland, Laura Gustafson, Tete Xiao, Spencer White- head, Alexander C Berg, Wan-Yen Lo, et al. Segment any- thing. InProceedings of the IEEE/CVF international confer- ence on computer vision, pages 4015–4026, 2023. 3

    work page 2023

  23. [23]

    PoseLib - Minimal Solvers for Camera Pose Estimation, 2020

    Viktor Larsson and contributors. PoseLib - Minimal Solvers for Camera Pose Estimation, 2020. 6

    work page 2020

  24. [24]

    Epnp: An accurate o(n) solution to the pnp problem.Inter- national journal of computer vision, 81(2):155–166, 2009

    Vincent Lepetit, Francesc Moreno-Noguer, and Pascal Fua. Epnp: An accurate o(n) solution to the pnp problem.Inter- national journal of computer vision, 81(2):155–166, 2009. 1, 5, 6 16

    work page 2009

  25. [25]

    Ground- ing image matching in 3d with mast3r

    Vincent Leroy, Yohann Cabon, and J´erˆome Revaud. Ground- ing image matching in 3d with mast3r. InEuropean Confer- ence on Computer Vision, pages 71–91. Springer, 2024. 2

    work page 2024

  26. [26]

    Hierarchical scene coordinate classification and regression for visual localization

    Xiaotian Li, Shuzhe Wang, Yi Zhao, Jakob Verbeek, and Juho Kannala. Hierarchical scene coordinate classification and regression for visual localization. InProceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pages 11983–11992, 2020. 2, 7, 12

    work page 2020

  27. [27]

    Learning neural volumetric pose features for camera localization

    Jingyu Lin, Jiaqi Gu, Bojian Wu, Lubin Fan, Renjie Chen, Ligang Liu, and Jieping Ye. Learning neural volumetric pose features for camera localization. InEuropean Conference on Computer Vision, pages 198–214. Springer, 2024. 2, 6, 7

    work page 2024

  28. [28]

    Hr-apr: Apr-agnostic framework with uncertainty estimation and hierarchical re- finement for camera relocalisation

    Changkun Liu, Shuai Chen, Yukun Zhao, Huajian Huang, Victor Prisacariu, and Tristan Braud. Hr-apr: Apr-agnostic framework with uncertainty estimation and hierarchical re- finement for camera relocalisation. In2024 IEEE Inter- national Conference on Robotics and Automation (ICRA), pages 8544–8550. IEEE, 2024. 6, 7

    work page 2024

  29. [29]

    GS-CPR: Efficient camera pose refinement via 3d gaussian splatting

    Changkun Liu, Shuai Chen, Yash Sanjay Bhalgat, Siyan HU, Ming Cheng, Zirui Wang, Victor Adrian Prisacariu, and Tris- tan Braud. GS-CPR: Efficient camera pose refinement via 3d gaussian splatting. InThe Thirteenth International Confer- ence on Learning Representations, 2025. 2, 6, 7, 11, 12

    work page 2025

  30. [30]

    Robust incremental structure-from-motion with hybrid fea- tures

    Shaohui Liu, Yidan Gao, Tianyi Zhang, R ´emi Pautrat, Jo- hannes L Sch ¨onberger, Viktor Larsson, and Marc Pollefeys. Robust incremental structure-from-motion with hybrid fea- tures. InEuropean Conference on Computer Vision, pages 249–269. Springer, 2024. 1, 2

    work page 2024

  31. [31]

    Distinctive image features from scale- invariant keypoints.International journal of computer vi- sion, 60(2):91–110, 2004

    David G Lowe. Distinctive image features from scale- invariant keypoints.International journal of computer vi- sion, 60(2):91–110, 2004. 2

    work page 2004

  32. [32]

    6dgs: 6d pose estimation from a single image and a 3d gaussian splatting model

    Bortolon Matteo, Theodore Tsesmelis, Stuart James, Fabio Poiesi, and Alessio Del Bue. 6dgs: 6d pose estimation from a single image and a 3d gaussian splatting model. InEuropean Conference on Computer Vision, pages 420–436. Springer,

    work page

  33. [33]

    Nerf: Representing scenes as neural radiance fields for view syn- thesis.Communications of the ACM, 65(1):99–106, 2021

    Ben Mildenhall, Pratul P Srinivasan, Matthew Tancik, Jonathan T Barron, Ravi Ramamoorthi, and Ren Ng. Nerf: Representing scenes as neural radiance fields for view syn- thesis.Communications of the ACM, 65(1):99–106, 2021. 1, 2

    work page 2021

  34. [34]

    Lens: Localization enhanced by nerf synthesis

    Arthur Moreau, Nathan Piasco, Dzmitry Tsishkou, Bogdan Stanciulescu, and Arnaud de La Fortelle. Lens: Localization enhanced by nerf synthesis. InConference on Robot Learn- ing, pages 1347–1356. PMLR, 2022. 2, 7

    work page 2022

  35. [35]

    Crossfire: Camera relocalization on self-supervised features from an implicit representation

    Arthur Moreau, Nathan Piasco, Moussab Bennehar, Dzmitry Tsishkou, Bogdan Stanciulescu, and Arnaud de La Fortelle. Crossfire: Camera relocalization on self-supervised features from an implicit representation. InProceedings of the IEEE/CVF International Conference on Computer Vision, pages 252–262, 2023. 2, 7

    work page 2023

  36. [36]

    Reassessing the limitations of cnn methods for camera pose regression

    Tony Ng, Adrian Lopez-Rodriguez, Vassileios Balntas, and Krystian Mikolajczyk. Reassessing the limitations of cnn methods for camera pose regression. InInternational Con- ference on 3D Vision, 2021. 7, 12

    work page 2021

  37. [37]

    Meshloc: Mesh-based visual localization

    V ojtech Panek, Zuzana Kukelova, and Torsten Sattler. Meshloc: Mesh-based visual localization. InEuropean Con- ference on Computer Vision, pages 589–609. Springer, 2022. 2

    work page 2022

  38. [38]

    Gaussian splatting feature fields for (privacy-preserving) vi- sual localization

    Maxime Pietrantoni, Gabriela Csurka, and Torsten Sattler. Gaussian splatting feature fields for (privacy-preserving) vi- sual localization. InProceedings of the Computer Vision and Pattern Recognition Conference, pages 1082–1092, 2025. 6, 7, 12

    work page 2025

  39. [39]

    Synthetic view generation for absolute pose regression and image syn- thesis

    Pulak Purkait, Cheng Zhao, and Christopher Zach. Synthetic view generation for absolute pose regression and image syn- thesis. InBMVC, page 69, 2018. 2

    work page 2018

  40. [40]

    Langsplat: 3d language gaussian splatting

    Minghan Qin, Wanhua Li, Jiawei Zhou, Haoqian Wang, and Hanspeter Pfister. Langsplat: 3d language gaussian splatting. InProceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pages 20051–20060, 2024. 2

    work page 2024

  41. [41]

    Learning transferable visual models from natural language supervi- sion

    Alec Radford, Jong Wook Kim, Chris Hallacy, Aditya Ramesh, Gabriel Goh, Sandhini Agarwal, Girish Sastry, Amanda Askell, Pamela Mishkin, Jack Clark, et al. Learning transferable visual models from natural language supervi- sion. InInternational conference on machine learning, pages 8748–8763. PmLR, 2021. 3

    work page 2021

  42. [42]

    R2d2: Reliable and repeatable detec- tor and descriptor.Advances in neural information process- ing systems, 32, 2019

    Jerome Revaud, Cesar De Souza, Martin Humenberger, and Philippe Weinzaepfel. R2d2: Reliable and repeatable detec- tor and descriptor.Advances in neural information process- ing systems, 32, 2019. 2

    work page 2019

  43. [43]

    From coarse to fine: Robust hierarchical localization at large scale

    Paul-Edouard Sarlin, Cesar Cadena, Roland Siegwart, and Marcin Dymczyk. From coarse to fine: Robust hierarchical localization at large scale. InProceedings of the IEEE/CVF conference on computer vision and pattern recognition, pages 12716–12725, 2019. 1, 2, 6, 7, 12

    work page 2019

  44. [44]

    Superglue: Learning feature matching with graph neural networks

    Paul-Edouard Sarlin, Daniel DeTone, Tomasz Malisiewicz, and Andrew Rabinovich. Superglue: Learning feature matching with graph neural networks. InProceedings of the IEEE/CVF conference on computer vision and pattern recognition, pages 4938–4947, 2020. 2

    work page 2020

  45. [45]

    Back to the feature: Learning robust camera localization from pixels to pose

    Paul-Edouard Sarlin, Ajaykumar Unagar, Mans Larsson, Hugo Germain, Carl Toft, Viktor Larsson, Marc Pollefeys, Vincent Lepetit, Lars Hammarstrand, Fredrik Kahl, et al. Back to the feature: Learning robust camera localization from pixels to pose. InProceedings of the IEEE/CVF con- ference on computer vision and pattern recognition, pages 3247–3257, 2021. 1, 2, 7

    work page 2021

  46. [46]

    Fast image- based localization using direct 2d-to-3d matching

    Torsten Sattler, Bastian Leibe, and Leif Kobbelt. Fast image- based localization using direct 2d-to-3d matching. In2011 International Conference on Computer Vision, pages 667–

    work page

  47. [47]

    Improving image-based localization by active correspondence search

    Torsten Sattler, Bastian Leibe, and Leif Kobbelt. Improving image-based localization by active correspondence search. InEuropean conference on computer vision, pages 752–765. Springer, 2012. 6, 7

    work page 2012

  48. [48]

    Structure- from-motion revisited

    Johannes L Schonberger and Jan-Michael Frahm. Structure- from-motion revisited. InProceedings of the IEEE con- ference on computer vision and pattern recognition, pages 4104–4113, 2016. 1

    work page 2016

  49. [49]

    Camera pose auto-encoders for improving pose regression

    Yoli Shavit and Yosi Keller. Camera pose auto-encoders for improving pose regression. InEuropean Conference on Computer Vision, pages 140–157. Springer, 2022. 1, 2 17

    work page 2022

  50. [50]

    Learning multi- scene absolute pose regression with transformers

    Yoli Shavit, Ron Ferens, and Yosi Keller. Learning multi- scene absolute pose regression with transformers. InPro- ceedings of the IEEE/CVF International Conference on Computer Vision, pages 2733–2742, 2021. 2

    work page 2021

  51. [51]

    Language embedded 3d gaussians for open- vocabulary scene understanding

    Jin-Chuan Shi, Miao Wang, Hao-Bin Duan, and Shao- Hua Guan. Language embedded 3d gaussians for open- vocabulary scene understanding. InProceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pages 5333–5343, 2024. 2

    work page 2024

  52. [52]

    Scene co- ordinate regression forests for camera relocalization in rgb-d images

    Jamie Shotton, Ben Glocker, Christopher Zach, Shahram Izadi, Antonio Criminisi, and Andrew Fitzgibbon. Scene co- ordinate regression forests for camera relocalization in rgb-d images. InProceedings of the IEEE conference on computer vision and pattern recognition, pages 2930–2937, 2013. 6

    work page 2013

  53. [53]

    Gsplatloc: Grounding key- point descriptors into 3d gaussian splatting for improved vi- sual localization.arXiv preprint arXiv:2409.16502, 2024

    Gennady Sidorov, Malik Mohrat, Denis Gridusov, Ruslan Rakhimov, and Sergey Kolyubin. Gsplatloc: Grounding key- point descriptors into 3d gaussian splatting for improved vi- sual localization.arXiv preprint arXiv:2409.16502, 2024. 1, 2, 3, 6, 7, 12

    work page arXiv 2024

  54. [54]

    Loftr: Detector-free local feature matching with transformers

    Jiaming Sun, Zehong Shen, Yuang Wang, Hujun Bao, and Xiaowei Zhou. Loftr: Detector-free local feature matching with transformers. InProceedings of the IEEE/CVF con- ference on computer vision and pattern recognition, pages 8922–8931, 2021. 2, 5

    work page 2021

  55. [55]

    icomma: Inverting 3d gaus- sian splatting for camera pose estimation via comparing and matching.arXiv preprint arXiv:2312.09031, 2023

    Yuan Sun, Xuan Wang, Yunfan Zhang, Jie Zhang, Caigui Jiang, Yu Guo, and Fei Wang. icomma: Inverting 3d gaus- sian splatting for camera pose estimation via comparing and matching.arXiv preprint arXiv:2312.09031, 2023. 2

    work page arXiv 2023

  56. [56]

    Neumap: Neural coordinate map- ping by auto-transdecoder for camera localization

    Shitao Tang, Sicong Tang, Andrea Tagliasacchi, Ping Tan, and Yasutaka Furukawa. Neumap: Neural coordinate map- ping by auto-transdecoder for camera localization. InPro- ceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pages 929–939, 2023. 1, 2, 6, 7

    work page 2023

  57. [57]

    24/7 place recognition by view synthesis

    Akihiko Torii, Relja Arandjelovic, Josef Sivic, Masatoshi Okutomi, and Tomas Pajdla. 24/7 place recognition by view synthesis. InProceedings of the IEEE conference on computer vision and pattern recognition, pages 1808–1817,

    work page

  58. [58]

    The unreasonable effectiveness of pre- trained features for camera pose refinement

    Gabriele Trivigno, Carlo Masone, Barbara Caputo, and Torsten Sattler. The unreasonable effectiveness of pre- trained features for camera pose refinement. InProceedings of the IEEE/CVF Conference on Computer Vision and Pat- tern Recognition, pages 12786–12798, 2024. 6, 7

    work page 2024

  59. [59]

    Learn- ing to navigate the energy landscape

    Julien Valentin, Angela Dai, Matthias Nießner, Pushmeet Kohli, Philip Torr, Shahram Izadi, and Cem Keskin. Learn- ing to navigate the energy landscape. In2016 Fourth In- ternational Conference on 3D Vision (3DV), pages 323–332. IEEE, 2016. 6

    work page 2016

  60. [60]

    Glace: Global local accelerated coordinate encoding

    Fangjinhua Wang, Xudong Jiang, Silvano Galliani, Christoph V ogel, and Marc Pollefeys. Glace: Global local accelerated coordinate encoding. InProceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pages 21562–21571, 2024. 1, 2, 6, 7, 12

    work page 2024

  61. [61]

    Dgc-gnn: Leveraging geometry and color cues for visual descriptor- free 2d-3d matching

    Shuzhe Wang, Juho Kannala, and Daniel Barath. Dgc-gnn: Leveraging geometry and color cues for visual descriptor- free 2d-3d matching. InProceedings of the IEEE/CVF Con- ference on Computer Vision and Pattern Recognition, pages 20881–20891, 2024. 1, 2

    work page 2024

  62. [62]

    Freesplat++: Generalizable 3d gaussian splatting for efficient indoor scene reconstruction, 2025

    Yunsong Wang, Tianxin Huang, Hanlin Chen, and Gim Hee Lee. Freesplat++: Generalizable 3d gaussian splatting for efficient indoor scene reconstruction.arXiv preprint arXiv:2503.22986, 2025. 13

    work page arXiv 2025

  63. [63]

    Ml-semreg: Boosting point cloud registration with multi-level semantic consistency

    Shaocheng Yan, Pengcheng Shi, and Jiayuan Li. Ml-semreg: Boosting point cloud registration with multi-level semantic consistency. InEuropean Conference on Computer Vision, pages 19–37. Springer, 2024. 5

    work page 2024

  64. [64]

    Turboreg: Turboclique for robust and efficient point cloud registration,

    Shaocheng Yan, Pengcheng Shi, Zhenjun Zhao, Kaixin Wang, Kuang Cao, Ji Wu, and Jiayuan Li. Turboreg: Turbo- clique for robust and efficient point cloud registration.arXiv preprint arXiv:2507.01439, 2025. 5

    work page arXiv 2025

  65. [65]

    Hemora: Unsupervised heuristic consensus sampling for ro- bust point cloud registration

    Shaocheng Yan, Yiming Wang, Kaiyan Zhao, Pengcheng Shi, Zhenjun Zhao, Yongjun Zhang, and Jiayuan Li. Hemora: Unsupervised heuristic consensus sampling for ro- bust point cloud registration. InProceedings of the Computer Vision and Pattern Recognition Conference, pages 1363– 1373, 2025. 4

    work page 2025

  66. [66]

    inerf: Inverting neural radiance fields for pose estimation

    Lin Yen-Chen, Pete Florence, Jonathan T Barron, Alberto Rodriguez, Phillip Isola, and Tsung-Yi Lin. inerf: Inverting neural radiance fields for pose estimation. In2021 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS), pages 1323–1330. IEEE, 2021. 1, 2

    work page 2021

  67. [67]

    Splatloc: 3d gaussian splatting-based visual localization for augmented reality.IEEE Transactions on Visualization and Computer Graphics, 2025

    Hongjia Zhai, Xiyu Zhang, Boming Zhao, Hai Li, Yijia He, Zhaopeng Cui, Hujun Bao, and Guofeng Zhang. Splatloc: 3d gaussian splatting-based visual localization for augmented reality.IEEE Transactions on Visualization and Computer Graphics, 2025. 2, 3, 7, 12

    work page 2025

  68. [68]

    Pnerfloc: Visual localization with point- based neural radiance fields

    Boming Zhao, Luwei Yang, Mao Mao, Hujun Bao, and Zhaopeng Cui. Pnerfloc: Visual localization with point- based neural radiance fields. InProceedings of the AAAI Conference on Artificial Intelligence, pages 7450–7459,

    work page

  69. [69]

    The nerfect match: Exploring nerf features for visual localization

    Qunjie Zhou, Maxim Maximov, Or Litany, and Laura Leal- Taix´e. The nerfect match: Exploring nerf features for visual localization. InEuropean Conference on Computer Vision, pages 108–127. Springer, 2024. 2, 6, 7, 12

    work page 2024

  70. [70]

    Feature 3dgs: Supercharging 3d gaussian splatting to enable distilled feature fields

    Shijie Zhou, Haoran Chang, Sicheng Jiang, Zhiwen Fan, Ze- hao Zhu, Dejia Xu, Pradyumna Chari, Suya You, Zhangyang Wang, and Achuta Kadambi. Feature 3dgs: Supercharging 3d gaussian splatting to enable distilled feature fields. InPro- ceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pages 21676–21685, 2024. 2, 4, 8, 9 18

    work page 2024