pith. sign in

arxiv: 2507.04503 · v2 · submitted 2025-07-06 · 💻 cs.CV · cs.RO

U-ViLAR: Uncertainty-Aware Visual Localization for Autonomous Driving via Differentiable Association and Registration

Xiaofan Li , Zhihao Xu , Chenming Wu , Zhao Yang , Yumeng Zhang , Jiang-Jiang Liu , Haibao Yu , Fan Duan
show 6 more authors
Xiaoqing Ye Yuan Wang Shirui Li Xun Sun Ji Wan Jun Wang
This is my paper

Pith reviewed 2026-05-19 05:44 UTC · model grok-4.3

classification 💻 cs.CV cs.RO
keywords visual localizationautonomous drivinguncertainty estimationbird's eye viewfeature associationmap registrationHD maps
0
0 comments X p. Extension

The pith

A visual localization system for autonomous driving guides map association and registration using estimates of perceptual and localization uncertainty.

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

The paper introduces U-ViLAR to perform accurate visual localization when GNSS signals are blocked by buildings and construction. It first maps camera features into bird's-eye-view space to align better with map data, then applies perceptual uncertainty to improve coarse feature association and localization uncertainty to refine precise registration. By combining these two steps the method claims to deliver robust performance that works with either high-definition maps or simpler navigation maps. Experiments are said to reach state-of-the-art results on standard benchmarks and to hold up during real-world testing on large autonomous-driving fleets in cities.

Core claim

By mapping visual features into BEV space and then using perceptual uncertainty to steer differentiable association while using localization uncertainty to steer differentiable registration, the framework balances large-scale coarse localization with fine-grained precise alignment, thereby achieving robust and accurate results with HD or navigation maps.

What carries the argument

Perceptual-uncertainty-guided association and localization-uncertainty-guided registration operating on BEV-projected visual features to align with map inputs.

If this is right

  • The combined association-plus-registration pipeline reaches state-of-the-art accuracy on multiple visual localization benchmarks.
  • The system adapts to either detailed HD maps or coarser navigation maps without retraining.
  • Real fleet deployments in challenging urban scenes maintain stable localization when GNSS is unavailable.
  • Uncertainty estimates directly reduce errors that would otherwise propagate from perception or from pose drift.

Where Pith is reading between the lines

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

  • Similar uncertainty-weighted matching could be applied to other map-relative tasks such as online map updating or semantic road segmentation.
  • If uncertainty maps prove reliable, the method might lower the frequency of map refreshes needed in rapidly changing construction zones.
  • Extending the same uncertainty guidance to multi-camera or camera-plus-LiDAR fusion would be a direct next test of the framework.

Load-bearing premise

Mapping camera features into bird's-eye view meaningfully improves spatial consistency with the map and the estimated uncertainties can steer association and registration without creating new errors or needing heavy parameter tuning.

What would settle it

A controlled ablation that removes either the BEV projection or the uncertainty weighting and shows no drop or an improvement in localization error on the same urban driving datasets would falsify the necessity of the proposed mechanisms.

Figures

Figures reproduced from arXiv: 2507.04503 by Chenming Wu, Fan Duan, Haibao Yu, Jiang-Jiang Liu, Ji Wan, Jun Wang, Shirui Li, Xiaofan Li, Xiaoqing Ye, Xun Sun, Yuan Wang, Yumeng Zhang, Zhao Yang, Zhihao Xu.

Figure 1
Figure 1. Figure 1: In Fine-grained Localization (left) and Large-scale Relocalization (right), U-ViLAR outperforms all existing meth￾ods across all metrics. Lat.’, Lon.’, and Ori.’ represent Lateral, Longitudinal, and Orientation, respectively, while R.’ denotes Re￾call. this context, visual localization using imprecise GNSS data is becoming increasingly critical. In autonomous driving systems, localization typically demands… view at source ↗
Figure 2
Figure 2. Figure 2: An overview of the proposed U-ViLAR. First, we extract features from image and map inputs, then align these features into BEV space to obtain visual and map features. These features undergo cross-modal fusion to enhance visual and map features (Sec. 3.1). During the Perceptual Uncertainty-guided (PU-Guided) Association phase (Sec. 3.2), we construct a similarity matrix from the features above and refine it… view at source ↗
Figure 4
Figure 4. Figure 4: Illustration of the Pose Registration. The solution space is constructed centered around the coarse pose, and the pre￾cise pose is obtained by estimating the probability of each candidate pose by the coarse prior. The similarity of positive sample pairs is optimized via a symmetric cross-entropy loss: \mathcal {L}_{\text {LM}} = \frac {1}{2\mathbf {K}} \sum _{i=1}^{\mathbf {K}} \left [ -\log \frac {e^{\mat… view at source ↗
Figure 5
Figure 5. Figure 5: Visualization on the nuScenes validation set. From left to right: (a) Front view, (b) Road-structure predictions overlaid with a BEV uncertainty heatmap, where red circles highlight re￾gions of high uncertainty, (c) Ground truth and (d) Inference results of the PU-aware Similarity Matrix [PITH_FULL_IMAGE:figures/full_fig_p008_5.png] view at source ↗
Figure 6
Figure 6. Figure 6: Trajectory segment from nuScenes validation set (left) and per-frame localization results (right). 𝐸𝑥, 𝐸𝑦, 𝐸𝜑: lateral, longitudinal, orientation errors; 𝑈𝑥, 𝑈𝑦, 𝑈𝜑: lateral, longitudinal, orientation uncertainties. representing the visual BEV space and map BEV space, respectively. In regions with poor association predictions (upper part), the Gaussian kernel is weaker and less accurate, indicating inferio… view at source ↗
read the original abstract

Accurate localization using visual information is a critical yet challenging task, especially in urban environments where nearby buildings and construction sites significantly degrade GNSS (Global Navigation Satellite System) signal quality. This issue underscores the importance of visual localization techniques in scenarios where GNSS signals are unreliable. This paper proposes U-ViLAR, a novel uncertainty-aware visual localization framework designed to address these challenges while enabling adaptive localization using high-definition (HD) maps or navigation maps. Specifically, our method first extracts features from the input visual data and maps them into Bird's-Eye-View (BEV) space to enhance spatial consistency with the map input. Subsequently, we introduce: a) Perceptual Uncertainty-guided Association, which mitigates errors caused by perception uncertainty, and b) Localization Uncertainty-guided Registration, which reduces errors introduced by localization uncertainty. By effectively balancing the coarse-grained large-scale localization capability of association with the fine-grained precise localization capability of registration, our approach achieves robust and accurate localization. Experimental results demonstrate that our method achieves state-of-the-art performance across multiple localization tasks. Furthermore, our model has undergone rigorous testing on large-scale autonomous driving fleets and has demonstrated stable performance in various challenging urban scenarios.

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

2 major / 1 minor

Summary. The paper proposes U-ViLAR, an uncertainty-aware visual localization framework for autonomous driving in GNSS-challenged urban environments. It extracts features from visual input, maps them into BEV space to improve spatial consistency with HD or navigation maps, introduces perceptual uncertainty-guided association to mitigate perception errors, and localization uncertainty-guided registration to reduce localization errors. By balancing the coarse-grained association with fine-grained registration, the method claims robust accurate localization, SOTA performance across tasks, and stable results from large-scale fleet testing in challenging scenarios.

Significance. If the uncertainty estimates are shown to be calibrated and the balancing mechanism yields measurable gains over standard differentiable matching, the work could meaningfully improve visual localization robustness for autonomous vehicles, particularly in fleet-scale urban deployments where perception and pose uncertainties are prevalent.

major comments (2)
  1. [Abstract] Abstract: the central claim that perceptual and localization uncertainties can be estimated from features and directly used to guide association/registration without introducing new error sources or dataset-specific tuning is load-bearing, yet the abstract provides no formulation, loss term, or calibration evidence (e.g., negative log-likelihood on held-out pose residuals) to support it.
  2. [Abstract] Abstract: the assertion of SOTA performance and fleet validation is presented without quantitative metrics, baselines, error bars, or ablation studies, preventing verification that the uncertainty-guided balancing actually improves upon standard association-plus-registration pipelines.
minor comments (1)
  1. [Abstract] Abstract: the phrasing 'introduced to mitigate errors' is vague; a brief statement of how the uncertainty heads are trained or regularized would improve clarity.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for the constructive feedback on our manuscript. We address each major comment point by point below, with proposed revisions to strengthen the presentation of our contributions in the abstract while preserving its conciseness.

read point-by-point responses

  1. Referee: [Abstract] Abstract: the central claim that perceptual and localization uncertainties can be estimated from features and directly used to guide association/registration without introducing new error sources or dataset-specific tuning is load-bearing, yet the abstract provides no formulation, loss term, or calibration evidence (e.g., negative log-likelihood on held-out pose residuals) to support it.

    Authors: We agree that the abstract's brevity limits its ability to convey the technical details of uncertainty estimation and guidance. The manuscript details the perceptual uncertainty estimation from BEV features and its use in differentiable association (Section 3.2), along with localization uncertainty-guided registration (Section 3.3), using a balancing mechanism that avoids dataset-specific tuning by relying on learned uncertainty maps. Supporting evidence, including performance gains and stability under uncertainty, appears in the experimental results. To address the concern, we will revise the abstract to briefly reference the uncertainty estimation from features and the guidance mechanisms without adding new error sources. revision: yes

  2. Referee: [Abstract] Abstract: the assertion of SOTA performance and fleet validation is presented without quantitative metrics, baselines, error bars, or ablation studies, preventing verification that the uncertainty-guided balancing actually improves upon standard association-plus-registration pipelines.

    Authors: The abstract summarizes the outcomes, while the full quantitative metrics, baselines, error bars, and ablation studies demonstrating the benefits of uncertainty-guided balancing over standard pipelines are provided in Section 4, including large-scale fleet testing results in challenging urban scenarios (Section 4.5). We acknowledge that incorporating brief quantitative context in the abstract would aid verification. In the revision, we will add concise references to the performance improvements and the role of the balancing mechanism while remaining within abstract length constraints. revision: yes

Circularity Check

0 steps flagged

No significant circularity; forward method pipeline with independent components

full rationale

The paper describes a sequential pipeline of feature extraction into BEV space followed by two introduced modules (perceptual uncertainty-guided association and localization uncertainty-guided registration) whose integration is presented as a design choice for balancing coarse and fine localization. No equations or claims reduce a performance metric or result by construction to a fitted parameter, self-citation chain, or renamed input; the uncertainty estimates are positioned as novel additions rather than derived from the target outputs. Experimental SOTA claims are treated as empirical validation outside the derivation itself, rendering the central framework self-contained against external benchmarks.

Axiom & Free-Parameter Ledger

0 free parameters · 1 axioms · 0 invented entities

Because only the abstract is available, the ledger is limited to high-level assumptions stated in the summary. The central claim depends on the effectiveness of BEV mapping and uncertainty guidance, which are presented as standard steps rather than derived results.

axioms (1)
  • domain assumption Mapping visual features to BEV space enhances spatial consistency with the map input
    Explicitly stated as the first processing step in the abstract.

pith-pipeline@v0.9.0 · 5783 in / 1230 out tokens · 54586 ms · 2026-05-19T05:44:24.627432+00:00 · methodology

discussion (0)

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

Lean theorems connected to this paper

Citations machine-checked in the Pith Canon. Every link opens the source theorem in the public Lean library.

What do these tags mean?
matches
The paper's claim is directly supported by a theorem in the formal canon.
supports
The theorem supports part of the paper's argument, but the paper may add assumptions or extra steps.
extends
The paper goes beyond the formal theorem; the theorem is a base layer rather than the whole result.
uses
The paper appears to rely on the theorem as machinery.
contradicts
The paper's claim conflicts with a theorem or certificate in the canon.
unclear
Pith found a possible connection, but the passage is too broad, indirect, or ambiguous to say the theorem truly supports the claim.

Reference graph

Works this paper leans on

49 extracted references · 49 canonical work pages · 1 internal anchor

  1. [1]

    Freak: Fast retina keypoint

    Alexandre Alahi, Raphael Ortiz, and Pierre Vandergheynst. Freak: Fast retina keypoint. In CVPR, pages 510–517, 2012. 1

    work page 2012

  2. [2]

    Learning to localize using a lidar intensity map

    Ioan Andrei B ˆarsan, Shenlong Wang, Andrei Pokrovsky, and Raquel Urtasun. Learning to localize using a lidar intensity map. 2018. 1

    work page 2018

  3. [3]

    Surf: Speeded up robust features

    Herbert Bay, Tinne Tuytelaars, and Luc Van Gool. Surf: Speeded up robust features. In ECCV, pages 404–417. Springer, 2006. 1

    work page 2006

  4. [4]

    Vizard: Reliable visual localization for autonomous vehicles in urban outdoor environments

    Mathias B¨ urki, Lukas Schaupp, Marcin Dymczyk, Renaud Dub´e, Cesar Cadena, Roland Siegwart, and Juan Nieto. Vizard: Reliable visual localization for autonomous vehicles in urban outdoor environments. pages 1124–1130. IEEE,

    work page

  5. [5]

    nuscenes: A mul- timodal dataset for autonomous driving

    Holger Caesar, Varun Bankiti, Alex H Lang, Sourabh Vora, Venice Erin Liong, Qiang Xu, Anush Krishnan, Yu Pan, Giancarlo Baldan, and Oscar Beijbom. nuscenes: A mul- timodal dataset for autonomous driving. In CVPR, pages 11621–11631, 2020. 6

    work page 2020

  6. [6]

    Brief: Binary robust independent elementary features

    Michael Calonder, Vincent Lepetit, Christoph Strecha, and Pascal Fua. Brief: Binary robust independent elementary features. In ECCV, pages 778–792. Springer, 2010. 1

    work page 2010

  7. [7]

    U-bev: Height- aware bird’s-eye-view segmentation and neural map-based relocalization

    Andrea Boscolo Camiletto, Alfredo Bochicchio, Alexander Liniger, Dengxin Dai, and Abel Gawel. U-bev: Height- aware bird’s-eye-view segmentation and neural map-based relocalization. 2024. 7

    work page 2024

  8. [8]

    I2d-loc: Camera localization via image to lidar depth flow

    Kuangyi Chen, Huai Yu, Wen Yang, Lei Yu, Sebastian Scherer, and Gui-Song Xia. I2d-loc: Camera localization via image to lidar depth flow. ISPRS Journal of Photogram- metry and Remote Sensing, 194:209–221, 2022. 3

    work page 2022

  9. [9]

    Superpoint: Self-supervised interest point detection and description

    Daniel DeTone, Tomasz Malisiewicz, and Andrew Rabi- novich. Superpoint: Self-supervised interest point detection and description. pages 224–236, 2018. 1, 2

    work page 2018

  10. [10]

    Uncertainty-aware vision-based metric cross-view geolocalization

    Florian Fervers, Sebastian Bullinger, Christoph Bodensteiner, Michael Arens, and Rainer Stiefelhagen. Uncertainty-aware vision-based metric cross-view geolocalization. In CVPR, pages 21621–21631, 2023. 2

    work page 2023

  11. [11]

    Vision meets robotics: The KITTI dataset

    Andreas Geiger, Philip Lenz, Christoph Stiller, and Raquel Urtasun. Vision meets robotics: The KITTI dataset. In CVPR, 2013. 6

    work page 2013

  12. [12]

    Gitnet: Geometric prior- based transformation for birds-eye-view segmentation

    Shi Gong, Xiaoqing Ye, Xiao Tan, Jingdong Wang, Errui Ding, Yu Zhou, and Xiang Bai. Gitnet: Geometric prior- based transformation for birds-eye-view segmentation. In ECCV, 2022. 2

    work page 2022

  13. [13]

    Openstreetmap: User- generated street maps

    Mordechai Haklay and Patrick Weber. Openstreetmap: User- generated street maps. IEEE Pervasive computing, 7(4):12– 18, 2008. 2, 3

    work page 2008

  14. [14]

    Zhang, Shaoqing Ren, and Jian Sun

    Kaiming He, X. Zhang, Shaoqing Ren, and Jian Sun. Deep residual learning for image recognition. Proceedings of the IEEE Conference on Computer Vision and Pattern Recogni- tion (CVPR), pages 770–778, 2016. 3

    work page 2016

  15. [15]

    Egovm: Achieving precise ego-localization using lightweight vectorized maps

    Yuzhe He, Shuang Liang, Xiaofei Rui, Chengying Cai, and Guowei Wan. Egovm: Achieving precise ego-localization using lightweight vectorized maps. 2024. 3

    work page 2024

  16. [16]

    BEVDet: High-performance Multi-camera 3D Object Detection in Bird-Eye-View

    Junjie Huang, Guan Huang, Zheng Zhu, and Dalong Du. Bevdet: High-performance multi-camera 3d object detection in bird-eye-view. arXiv preprint arXiv:2112.11790, 2021. 2

    work page internal anchor Pith review Pith/arXiv arXiv 2021

  17. [17]

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

    Alex Kendall, Matthew Grimes, and Roberto Cipolla. Posenet: A convolutional network for real-time 6-dof camera relocalization. In ICCV, 2015. 2

    work page 2015

  18. [18]

    Slicematch: Geometry-guided aggregation for cross-view pose estimation

    Ted Lentsch, Zimin Xia, Holger Caesar, and Julian FP Kooij. Slicematch: Geometry-guided aggregation for cross-view pose estimation. In CVPR, pages 17225–17234, 2023. 2

    work page 2023

  19. [19]

    Map-based precision vehicle localization in urban environ- ments

    Jesse Levinson, Michael Montemerlo, and Sebastian Thrun. Map-based precision vehicle localization in urban environ- ments. volume 4, page 1, 2007. 2

    work page 2007

  20. [20]

    Robust vehicle localiza- tion in urban environments using probabilistic maps

    Jesse Levinson and Sebastian Thrun. Robust vehicle localiza- tion in urban environments using probabilistic maps. pages 4372–4378, May 2010. 2

    work page 2010

  21. [21]

    Bevdepth: Acquisition of reliable depth for multi-view 3d object detec- tion

    Yinhao Li, Zheng Ge, Guanyi Yu, Jinrong Yang, Zengran Wang, Yukang Shi, Jianjian Sun, and Zeming Li. Bevdepth: Acquisition of reliable depth for multi-view 3d object detec- tion. In AAAI, volume 37, pages 1477–1485, 2023. 2

    work page 2023

  22. [22]

    Worldwide pose estimation using 3d point clouds

    Yunpeng Li, Noah Snavely, Daniel P Huttenlocher, and Pas- cal Fua. Worldwide pose estimation using 3d point clouds. Large-Scale Visual Geo-Localization, pages 147–163, 2016. 2

    work page 2016

  23. [23]

    Bevformer: Learn- ing bird’s-eye-view representation from multi-camera images via spatiotemporal transformers

    Zhiqi Li, Wenhai Wang, Hongyang Li, Enze Xie, Chonghao Sima, Tong Lu, Yu Qiao, and Jifeng Dai. Bevformer: Learn- ing bird’s-eye-view representation from multi-camera images via spatiotemporal transformers. In Computer Vision–ECCV 2022: 17th European Conference, Tel Aviv, Israel, October 23–27, 2022, Proceedings, Part IX , pages 1–18. Springer,

    work page 2022

  24. [24]

    Pixel-perfect structure-from-motion with featuremetric refinement

    Philipp Lindenberger, Paul-Edouard Sarlin, Viktor Larsson, and Marc Pollefeys. Pixel-perfect structure-from-motion with featuremetric refinement. In ICCV, pages 5987–5997, 2021. 2

    work page 2021

  25. [25]

    Work smart, not hard: Recalling relevant experiences for vast-scale but time-constrained localisation

    Chris Linegar, Winston Churchill, and Paul Newman. Work smart, not hard: Recalling relevant experiences for vast-scale but time-constrained localisation. pages 90–97, 2015. 1

    work page 2015

  26. [26]

    Gift: A real-time and scalable 3d spatial feature de- scriptor

    Yong Liu, Lingjing Fan, Sen Xiang, Jiaqi Pan, Mingliang Cheng, Jinhui Tang, Fei Zhao, Zilei Wang, and Qingxiong Zhao. Gift: A real-time and scalable 3d spatial feature de- scriptor. In CVPR, pages 6783–6791, 2017. 1 9

    work page 2017

  27. [27]

    Petr: Position embedding transformation for multi-view 3d object detection

    Yingfei Liu, Tiancai Wang, Xiangyu Zhang, and Jian Sun. Petr: Position embedding transformation for multi-view 3d object detection. In ECCV, pages 531–548, 2022. 2

    work page 2022

  28. [28]

    Vectormapnet: End-to-end vectorized hd map learning

    Yicheng Liu, Tianyuan Yuan, Yue Wang, Yilun Wang, and Hang Zhao. Vectormapnet: End-to-end vectorized hd map learning. pages 22352–22369, 2023. 2

    work page 2023

  29. [29]

    Distinctive image features from scale- invariant keypoints

    David G Lowe. Distinctive image features from scale- invariant keypoints. IJCV, 60(2):91–110, 2004. 1

    work page 2004

  30. [30]

    Automated map reading: image based localisation in 2-d maps using binary semantic descriptors

    Pilailuck Panphattarasap and Andrew Calway. Automated map reading: image based localisation in 2-d maps using binary semantic descriptors. pages 6341–6348, 2018. 2

    work page 2018

  31. [31]

    Monocular localization in hd maps by combining semantic segmentation and distance transform

    Jan-Hendrik Pauls, K¨ ursat Petek, Fabian Poggenhans, and Christoph Stiller. Monocular localization in hd maps by combining semantic segmentation and distance transform. In 2020 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS), pages 4595–4601. IEEE, 2020. 2

    work page 2020

  32. [32]

    Lift, splat, shoot: Encoding images from arbitrary camera rigs by implicitly unprojecting to 3d

    Jonah Philion and Sanja Fidler. Lift, splat, shoot: Encoding images from arbitrary camera rigs by implicitly unprojecting to 3d. In ECCV, pages 194–210, 2020. 2

    work page 2020

  33. [33]

    A sim2real deep learning approach for the transformation of images from multiple vehicle-mounted cameras to a seman- tically segmented image in bird’s eye view

    Lennart Reiher, Bastian Lampe, and Lutz Eckstein. A sim2real deep learning approach for the transformation of images from multiple vehicle-mounted cameras to a seman- tically segmented image in bird’s eye view. In International Conference on Intelligent Transportation Systems (ITSC) , pages 1–7, 2020. 2

    work page 2020

  34. [34]

    R2d2: Reliable and repeatable detectors and descriptors

    Jerome Revaud, Philippe Weinzaepfel, Ce ´esar de Souza, Noe´e Pion, Gabriela Csurka, Yann Cabon, and Martin Hu- menberger. R2d2: Reliable and repeatable detectors and descriptors. In Advances in Neural Information Processing Systems (NeurIPS), pages 12405–12415, 2019. 1, 2

    work page 2019

  35. [35]

    U- net: Convolutional networks for biomedical image segmen- tation

    Olaf Ronneberger, Philipp Fischer, and Thomas Brox. U- net: Convolutional networks for biomedical image segmen- tation. In Medical Image Computing and Computer-Assisted Intervention–MICCAI, pages 234–241, 2015. 3

    work page 2015

  36. [36]

    Orb: An efficient alternative to sift or surf

    Ethan Rublee, Vincent Rabaud, Kurt Konolige, and Gary Bradski. Orb: An efficient alternative to sift or surf. In 2011 International conference on computer vision , pages 2564–

    work page 2011

  37. [37]

    You are here: Geolocation by embedding maps and images

    Noe Samano, Mengjie Zhou, and Andrew Calway. You are here: Geolocation by embedding maps and images. InECCV, pages 502–518. Springer, 2020. 2

    work page 2020

  38. [38]

    Superglue: Learning feature match- ing with graph neural networks

    Paul-Edouard Sarlin, Daniel DeTone, Tomasz Malisiewicz, and Andrew Rabinovich. Superglue: Learning feature match- ing with graph neural networks. InCVPR, pages 4938–4947,

    work page

  39. [39]

    Orienternet: Visual localization in 2d public maps with neural matching

    Paul-Edouard Sarlin, Daniel DeTone, Tsun-Yi Yang, Armen Avetisyan, Julian Straub, Tomasz Malisiewicz, Samuel Rota Bul`o, Richard Newcombe, Peter Kontschieder, and Vasileios Balntas. Orienternet: Visual localization in 2d public maps with neural matching. In CVPR, pages 21632–21642, 2023. 2, 3

    work page 2023

  40. [40]

    Large- scale direct monocular slam

    Torsten Sattler, Laura Leal-Taix´e, and Marc Pollefeys. Large- scale direct monocular slam. In Proceedings of the IEEE International Conference on Computer Vision (ICCV), 2017. 2

    work page 2017

  41. [41]

    Beyond cross-view image retrieval: Highly accurate vehicle localization using satellite image

    Yujiao Shi and Hongdong Li. Beyond cross-view image retrieval: Highly accurate vehicle localization using satellite image. In CVPR, pages 17010–17020, 2022. 6

    work page 2022

  42. [42]

    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. In Proceedings of the IEEE conference on computer vision and pattern recognition, pages 2930–2937, 2013. 2

    work page 2013

  43. [43]

    Sensor fusion-based low-cost vehicle localization system for complex urban environments

    Jae Kyu Suhr, Jeungin Jang, Daehong Min, and Ho Gi Jung. Sensor fusion-based low-cost vehicle localization system for complex urban environments. 18(5):1078–1086, 2016. 2

    work page 2016

  44. [44]

    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. In CVPR, pages 8922–8931, 2021. 2

    work page 2021

  45. [45]

    Visual semantic localization based on hd map for autonomous vehicles in urban scenarios

    Huayou Wang, Changliang Xue, Yanxing Zhou, Feng Wen, and Hongbo Zhang. Visual semantic localization based on hd map for autonomous vehicles in urban scenarios. pages 11255–11261, 2021. 2

    work page 2021

  46. [46]

    Detr3d: 3d object detection from multi-view images via 3d-to-2d queries

    Yue Wang, Vitor Campagnolo Guizilini, Tianyuan Zhang, Yilun Wang, Hang Zhao, and Justin Solomon. Detr3d: 3d object detection from multi-view images via 3d-to-2d queries. pages 180–191, 2022. 2

    work page 2022

  47. [47]

    Maplocnet: Coarse-to-fine feature registration for visual re-localization in navigation maps

    Hang Wu, Zhenghao Zhang, Siyuan Lin, Xiangru Mu, Qiang Zhao, Ming Yang, and Tong Qin. Maplocnet: Coarse-to-fine feature registration for visual re-localization in navigation maps. pages 13198–13205. IEEE, 2024. 2

    work page 2024

  48. [48]

    Uncertainty-aware learning for video frame interpola- tion

    Haoqiang Zhang, Wenhan Luo, Wenxiu Yu, and Xiaoyong Yang. Uncertainty-aware learning for video frame interpola- tion. pages 10543–10554, 2022. 4

    work page 2022

  49. [49]

    Bev-locator: An end-to-end visual semantic localization network using multi-view im- ages

    Zhihuang Zhang, Meng Xu, Wenqiang Zhou, Tao Peng, Liang Li, and Stefan Poslad. Bev-locator: An end-to-end visual semantic localization network using multi-view im- ages. In ECCV, 2022. 2, 3 10

    work page 2022