Pith. sign in

REVIEW 4 major objections 5 minor 48 references

Topometric Autonomous Vehicle Localization by Combining Visual Embeddings and Feed-Forward 3D Models

T0 review · 4 major / 5 minor · reviewed 2026-08-07 · deepseek-v4-flash

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗ Review JSON

Pith's one-line read Conditioning feed-forward 3D pose estimation on a particle filter's belief over places makes sequential appearance-based localization metrically accurate, with compact maps and lower mean and P90 position errors than appearance-only…

desk verdict A genuinely new integration of VPR and feed-forward 3D geometry into a topometric particle filter, with solid results and honest limitations—but thin statistics and test-set tuning keep it short of a clean accept. read the letter →

arxiv 2608.06021 v1 pith:AAPGIZC6 submitted 2026-08-06 cs.RO

Eulogio Quemada-Torres , Alberto Jaenal , Francisco-Angel Moreno , Javier Gonzalez-Jimenez This is my paper · ORCID
classification cs.RO
keywords visuallocalizationplacerecognitiontopometricmapsMonteCarlofeed-forward3Dmodelsparticlefilterneuralmetricposeestimationappearance-based
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

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

The reading

Visual place recognition is compact and robust to weather and lighting change, but its pose estimates are too coarse for precise localization, while feed-forward 3D geometry models are accurate but costly on large image sets. This paper claims a topometric framework—topological place tracking fused with metric poses—resolves the trade-off: a particle filter keeps a belief over places, that belief selects a small set of map references for a feed-forward 3D pose network, and the resulting metric observation is fused with the place-recognition likelihood. Offline, the map is built automatically with HDBSCAN clustering and farthest-point representative selection, so no cluster count is preset. On COLD, 4Seasons, and RobotCar, the authors report the lowest mean and P90 position errors on every dataset relative to their appearance-based baselines, the best joint position–yaw recall on COLD and RobotCar, and deployable maps roughly five to over sixty times smaller than dense image maps. Descriptor extractor and 3D model remain interchangeable, which the paper treats as a design feature.

What carries the argument

The load-bearing mechanism is belief-conditioned neural metric refinement inside a topometric particle filter. Each particle carries a pose and an associated topological place; the aggregated weights define a belief over places, and candidate clusters are kept when their belief is within a relative threshold of the dominant cluster. From those clusters, a compact reference set N_t is selected by descriptor distance with explicit lower and upper bounds, then processed by the feed-forward 3D model g (default Depth Anything 3) together with the query image. The model's internal-frame poses are aligned to the map frame by a robust Sim(2) fit, and the resulting observation is evaluated against each particle with a Gaussian SE(2) distance (Equation 18), fused with the appearance likelihood through normalized log-likelihoods and mixing weight α. The offline map is built by clustering standardized pose-feature vectors with HDBSCAN, compressing each cluster into pose/descriptor means and a block-diagonal covariance, and keeping representative images by farthest-point sampling in descriptor space.

What would settle it

On the 4Seasons frames with the highest position errors, measure the pairwise map distance among the belief-selected reference images; if P90 error rises sharply whenever those references span a large area, the single-Sim(2) alignment assumption is confirmed as the failure point, and if it does not, the system is more robust to inconsistent references than the paper's own caveat suggests.

Watch

Extended reading notes

Core claim

On its own terms, the discovery is that the spatial belief of a particle filter is a better selector of evidence than the raw query image: it decides which reference images the metric estimator sees, and therefore which metric hypothesis is even formed. The metric observation is computed by taking the belief-selected candidate places, choosing up to K representative map images by descriptor distance, running the feed-forward 3D model jointly on the query and those references, and aligning the predicted poses to the georeferenced map with a robust Sim(2) fit (Least Median of Squares over Umeyama fits). That alignment yields a planar pose observation that enters the filter as a metric likelihood and is fused in log-space with the appearance likelihood. The paper reports that the resulting system obtains the lowest mean and P90 position errors on all three benchmarks (for example, mean position error of 0.124 m on COLD, 0.268 m on 4Seasons, and 0.494 m on RobotCar), and that the temporal belief prevents large failures by rejecting retrieval matches that disagree with the tracked region.

Load-bearing premise

The system assumes the handful of reference images chosen for a metric observation are spatially consistent enough that one alignment can place them into the map; in large or elongated places an inconsistent reference set can inject a wrong pose into the filter, an effect the authors say they observed on 4Seasons.

Editorial extensions

If this is right

  • Sequential VPR localization can reach metric accuracy close to feature-based pipelines without storing a dense image map, because each metric observation needs only a handful of belief-selected references.
  • Large localization failures under perceptual aliasing are suppressed: the belief restricts evidence to spatially plausible places, and the COLD case study shows the tracked mode surviving where direct retrieval follows a visually similar but wrong region.
  • Deployable map storage drops by factors of roughly 5 to more than 60 relative to dense VPR+FF3D maps, easing long-term and large-scale deployment.
  • The system is tunable along several axes: swapping the descriptor or the 3D model changes the accuracy–latency balance, particle count saturates around 200, fusion is best at α = 0.7, and the FPS threshold trades map size against accuracy.
  • Appearance and metric evidence are complementary; neither the appearance-only nor the metric-only update matches the fused update on the three-dataset average.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

  • The same belief-conditioning recipe should transfer to any metric pose estimator whose cost grows with input count, such as local-feature pose solvers or dense relocalizers; the paper only demonstrates it for feed-forward 3D models, so this is an extension, not a claim.
  • Because the paper acknowledges that fixed fusion ignores FF3D confidence, an uncertainty-aware weighting that down-weights the metric term in elongated or ambiguous places is a natural next step and would address exactly the 4Seasons failure mode it reports.
  • A direct test of the spatial-consistency assumption: measure per-place P90 position error against place length or descriptor diversity; the paper's own caveat predicts higher error tails precisely where references can span inconsistent locations.
  • Automatic place discovery by HDBSCAN suggests a path to incremental map updates from repeated traversals, since no predefined number of places is required, although the paper assumes a static georeferenced map.
Share X Bluesky LinkedIn Reddit HN

Signed reviews

No signed human review yet.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Referee Report

4 major / 5 minor

Summary. The paper presents a topometric visual localization framework that combines probabilistic visual place recognition (VPR) with feed-forward neural 3D geometry (FF3D) models. The offline stage automatically abstracts a dense georeferenced image map into probabilistic places using HDBSCAN and selects representative images via farthest-point sampling. The online stage runs a particle filter that propagates odometry, conditions reference selection on the belief over places, obtains a metric pose observation by aligning FF3D predictions to the map frame with a robust Sim(2) fit, and fuses appearance and metric likelihoods. The method is evaluated on COLD, 4Seasons, and RobotCar, reporting lower mean and P90 position errors than the appearance-based ALLOM baseline and a direct VPR+FF3D retrieval baseline, with substantially reduced map size. The paper also includes ablations over descriptors, FF3D models, particle count, fusion weight, and map compression, plus a focused perceptual-aliasing case study.

Significance. If the reported results hold, the framework is a useful engineering contribution: it demonstrates a modular way to combine compact appearance-based mapping with neural metric estimation, and the ablations over many descriptor and FF3D backends are unusually thorough. The deployable-map-size reduction relative to dense image maps is practically relevant. The perceptual-aliasing case study is a nice attempt to isolate the mechanism by which sequential belief helps. However, the central comparative claims are currently supported only by point estimates without variance or significance testing, and the paper's own acknowledged failure mode on 4Seasons directly concerns the mechanism that is supposed to deliver tail robustness. The contribution is therefore promising but not yet established at the level claimed.

major comments (4)
  1. [Sec. 4.3, Table 2] The paper reports only three random seeds per query sequence and explicitly states 'We do not report standard deviations or make statistical-significance claims.' Given that the headline claims are 'substantially improves' and 'lowest mean and P90 position errors on every dataset,' the absence of any variance measure or significance test is load-bearing. For example, on 4Seasons the AUC gap between Ours (0.864) and VPR+FF3D (0.897) is nontrivial and in the opposite direction; without run-to-run variance, the reader cannot tell whether the reported mean/P90 differences are within the noise of the stochastic filter. The authors should report per-seed results, standard deviations or confidence intervals, and at least a paired test across the three seeds per sequence.
  2. [Sec. 3.2.3, Eq. (18), Algorithm 2, Conclusion] The metric observation x^g_t is produced by a single robust Sim(2) fit that assumes the selected reference images are geometrically consistent in the map frame. Algorithm 2 selects references by descriptor diversity and cluster membership but does not enforce mutual pose consistency. The paper itself acknowledges in the Conclusion that 'Large or elongated places may admit inconsistent references and allow erroneous metric hypotheses to influence the belief, as observed on 4Seasons.' Because the claimed advantage over direct VPR+FF3D is precisely the prevention of large failures, this failure mode strikes at the central claim. The manuscript should quantify how often the Sim(2) consensus is corrupted (per dataset), and either add a consistency check that rejects or downweights unreliable metric observations or explicitly report the extent to which the final tail-error improvements depend on filtering out such cases.
  3. [Sec. 4.5, Figures 3–6, Table 3] The default configuration (MixVPR-512, DA3-Large, Np=200, alpha=0.7, tau_diversity=0.075, K=9, d_min/d_max) is selected by maximizing AUC_avg on the same three datasets that are later used for the main comparison in Table 2. This creates a selection bias: the reported performance of Ours is the performance of the best configuration chosen on these datasets, whereas VPR+FF3D and ALLOM are not tuned in the same way. Without a nested validation split or a held-out dataset for final evaluation, the magnitude of the improvement over the baselines is likely optimistic. The authors should either use a separate validation set for configuration selection or report the performance of a configuration chosen without dataset-specific tuning.
  4. [Table 2, 4Seasons row; Conclusion] The conclusion states the system 'obtains the lowest mean and P90 position errors on every dataset,' which is true in Table 2, but the broader claim that it 'substantially improves' over appearance-based topometric methods is weakened by the 4Seasons results: Ours has higher median position error and yaw errors than VPR+FF3D, and lower AUC. The text acknowledges this, but the abstract and conclusion still foreground the uniform-improvement narrative. The authors should either temper the claim to emphasize mean/P90 tail metrics, or provide a principled argument for why those metrics are the correct primary criterion, rather than the AUC that appears in the ablation section as the main selection metric.
minor comments (5)
  1. [Table 2] Several numeric entries appear to lack separating spaces, for example '4.1161.2087.398' and '0.2680.1730.558'; these should read '4.116 1.208 7.398' and '0.268 0.173 0.558' respectively.
  2. [Sec. 4.2, Table 1] The motion-noise standard deviations are listed as 'Indoor(0.025,0.010,0.015)' and 'Outdoor(0.25,0.10,0.015)' without explicit units; adding '[m, m, rad]' in the table header would remove ambiguity.
  3. [Eq. (18)] The metric d_SE(2) is described as adding Euclidean translation error in meters and wrapped yaw error in radians 'with unit rotational weight.' This is not a standard geodesic distance on SE(2) unless the weight is explicitly defined; please clarify that the distance is a weighted sum with an implicit scale factor of 1 m/rad and state this in the text.
  4. [Algorithm 2] Step 6 defines the valid set V as candidates with d_min <= delta_i <= d_max, but Step 8 then selects the member of V with the smallest delta_i, which is by construction the one closest to d_min. This makes the upper bound d_max only a filter on the maximum diversity; the selection criterion is effectively nearest-descriptor among those above d_min. The authors should explain the intended role of d_max more clearly.
  5. [General] The notation for the particle count is inconsistent: Eq. (8) uses N_p as a subscript, while Table 1 and the text use 'Np'. Please unify the notation.

Circularity Check

0 steps flagged · score 0.0 of 10

No circularity: the metric observation is a standard feed-forward 3D alignment, results are benchmarked on external datasets, and self-citations are background rather than load-bearing.

full rationale

The paper's derivation chain is self-contained. The FF3D metric observation x^g_t (Sec. 3.2.3, Eqs. 16-17) is produced by running a pretrained feed-forward geometry model on the query plus selected references and aligning the predicted reference poses to known map poses via a robust Sim(2) fit; the query pose estimate is not defined in terms of the filter output or the claimed accuracy result. The particle-filter update (Eqs. 18-23) combines this independent observation with an appearance likelihood that reuses the authors' earlier Jaenal et al. (2022) model, but that reuse is background methodology, not a self-citation invoked to force the paper's conclusion. The central accuracy claims are evaluated on three external benchmarks (COLD, 4Seasons, RobotCar) with query sequences distinct from the traversals used to build the maps, so the reported numbers are not produced by construction. The ablations in Sec. 4.5 select hyperparameters (descriptor, FF3D model, N_p, alpha, tau_diversity) on the same datasets; this is a potential selection-bias or overfitting concern, but it is not circularity under the required definition because no fitted parameter is renamed as a prediction and the paper explicitly reports no statistical-significance claims. The acknowledged 4Seasons limitation about large or elongated places admitting inconsistent references is a robustness failure mode of the geometric-consistency assumption, not a circular step. Self-citations to Jaenal et al. (2022, 2023) provide the map and filtering backbone, but they are not load-bearing as a substitute for evidence; the new FF3D metric refinement and the external benchmark evaluation stand independently.

Assumptions & free parameters 6 free parameters · 6 assumptions · 0 invented entities

The central claim depends on several domain assumptions about planar motion, descriptor distributions, and FF3D alignment quality. The main free parameters are the likelihood fusion weight, metric noise, particle count, and map compression threshold, all tuned on the same three datasets used for final evaluation.

free parameters (6)
  • alpha (likelihood fusion weight) = 0.70
    Selected by maximizing AUC_avg across the same COLD/4Seasons/RobotCar datasets (Fig. 5).
  • sigma_g (metric likelihood noise) = 0.20
    Chosen as part of default configuration (Table 1); affects the metric likelihood in Eq. 18.
  • N_p (particle count) = 200
    Ablation in Fig. 4 selects 200 as a balance between AUC and latency on the evaluation datasets.
  • tau_diversity (FPS stopping threshold) = 0.075
    Map compression threshold picked from Fig. 6 to keep AUC close to the unreduced baseline across all three datasets.
  • K (max reference images) = 9
    Fixed number of references for all experiments; also used by the VPR+FF3D baseline.
  • Descriptor and FF3D model choices = MixVPR-512, DA3-Large
    Selected from ablations in Fig. 3 and Table 3 as balanced operating points on the same datasets.
assumptions (6)
  • domain assumption Planar SE(2) motion and available odometry.
    The particle filter propagates poses in SE(2) using odometry u_t (Eq. 12), and FF3D transforms are projected to planar components. This limits applicability to ground vehicles. Location: Sec 3.2.2, Sec 4.2.
  • domain assumption Conditional independence between pose and appearance within a place.
    Eq. 7 uses a block-diagonal pose-descriptor covariance with zero off-diagonal blocks, assuming pose and appearance are independent per place. Location: Sec 3.1.2.
  • domain assumption Isotropic Gaussian descriptor distribution per place.
    The descriptor covariance is summarized by scalar sigma^2_{j,d}, and the appearance likelihood P(d^q|C_j) is computed from this distribution, following Jaenal et al. 2022. Location: Sec 3.1.2 and Sec 3.2.4.
  • domain assumption FF3D alignment via Sim(2) yields accurate query pose.
    The metric observation x^g_t is obtained by aligning FF3D predictions to map poses with LMedS+Umeyama (Sec 3.2.3). If references are inconsistent, the fit breaks down; the authors acknowledge this in the conclusion.
  • domain assumption HDBSCAN partitions the map into spatially and heading-coherent places.
    Clustering is performed on standardized 4D pose features (Eqs. 4-5); this assumes Euclidean distance in that space reflects topological place structure. Location: Sec 3.1.1.
  • standard math Standard particle filter and resampling theory.
    The framework relies on standard Monte Carlo localization assumptions about importance weights, propagation, and resampling (Eqs. 8-24). Location: Sec 3.2.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Topometric Autonomous Vehicle Localization by Combining Visual Embeddings and Feed-Forward 3D Models." pith.science (2026). https://pith.science/paper/AAPGIZC6

@misc{pith2026260806021,
  author       = {Pith},
  title        = {Pith review of: Topometric Autonomous Vehicle Localization by Combining Visual Embeddings and Feed-Forward 3D Models},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/AAPGIZC6}},
  note         = {Machine review of arXiv:2608.06021}
}
read the original abstract

Effective Visual Localization (VL) requires a map of the environment that combines compactness for efficient scalability with robustness against visual appearance changes and metric precision. Through low-dimensional image embeddings, Visual Place Recognition (VPR) is able to successfully meet the first two requirements, but its low metric accuracy makes it less suitable than standard VL approaches based on local features or neural representations. This limitation can be overcome by integrating VPR with the accurate local trajectory estimates produced by feed-forward neural 3D geometry (FF3D) models. In this paper, we address sequential appearance-based localization through a topometric framework that iteratively combines probabilistic VPR with FF3D metric pose estimation in controlled image sets. Our approach proposes an automatic offline mapping tool that models the topometric pose-appearance interaction in the different parts of the scene. This map is later employed by an online particle filter that estimates the pose from odometry and belief over places for FF3D inference, successfully incorporating neural metric estimation into probabilistic appearance-based localization. We extensively evaluate the framework on three known benchmarks, demonstrating substantial improvements over existing appearance-based methods. The modularity of our approach allows the descriptor extractor and FF3D model to remain interchangeable, and a focused analysis further shows that sequential belief can mitigate severe failures under perceptual aliasing.

Figures

Figures reproduced from arXiv: 2608.06021 by the authors.

Figure 1
Figure 1. Overview of the proposed topometric localization pipeline. The offline stage builds an abstracted appearance map from georeferenced images by constructing topological places and selecting representative samples. The online stage uses the resulting map in a topometric particle filter that combines appearance-based weighting with cluster-conditioned neural pose estimation. Formally, let the input map be M = {mi} N i=1… view at source ↗
Figure 2
Figure 2. Topometric maps constructed from the three reference sets. Retained reference samples are grouped by assigned place, ellipses show the planar place covariance, and arrows indicate the mean heading. Best in color [PITH_FULL_IMAGE:figures/full_fig_p008_2.png] view at source ↗
Figure 3
Figure 3. Visual-descriptor ablation. Lines connect dimensional variants of the same descriptor family, while individual markers denote architectures evaluated at one dimensionality. The enlarged marker identifies the selected configuration. dimensions, EigenPlaces at four sizes over the same range, and MixVPR at three sizes ranging from 128 to 4,096 dimensions. The remaining descriptor families are evaluated at their standar… view at source ↗
Figures from the paper (4 more)
Figure 5
Figure 5. Figure 5: AUCavg for different values of the parameter α, which mixes appearance and metric sources. The highlighted marker identifies the selected value α = 0.7 [PITH_FULL_IMAGE:figures/full_fig_p010_5.png]
Figure 4
Figure 4. Figure 4: AUCavg and particle-filter latency as Np increases. The enlarged marker identifies the selected value Np = 200 [PITH_FULL_IMAGE:figures/full_fig_p010_4.png]
Figure 7
Figure 7. Figure 7: Reference selection at COLD frame 843. The values above each image give its ground-truth planar pose, with position in meters and yaw in degrees. Our method selects a reference in the tracked region, whereas direct VPR+FF3D retrieves a visually similar image from an in…
Figure 8
Figure 8. Figure 8: Position error during the selected COLD perceptual-aliasing case. Both methods process the complete query sequence from its beginning; the curve for Ours is the mean across seeds. The qualitative example at frame 843 in [PITH_FULL_IMAGE:figures/full_fig_p011_8.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

48 extracted references · 45 canonical work pages

  1. [1]

    IEEE Robotics and Automation Letters , year =

    Jaenal, Alberto and Moreno, Francisco-Angel and Gonzalez-Jimenez, Javier , title =. IEEE Robotics and Automation Letters , year =

    work page

  2. [2]

    The International Journal of Robotics Research , year =

    Jaenal, Alberto and Moreno, Francisco-Angel and Gonzalez-Jimenez, Javier , title =. The International Journal of Robotics Research , year =

    work page

  3. [3]

    Toward a Unified Bayesian Approach to Hybrid Metric--Topological

    Blanco, Jose-Luis and Fern. Toward a Unified Bayesian Approach to Hybrid Metric--Topological. IEEE Transactions on Robotics , year =

    work page

  4. [4]

    Campello, Ricardo J. G. B. and Moulavi, Davoud and Sander, J. Density-Based Clustering Based on Hierarchical Density Estimates , booktitle =. 2013 , pages =

    2013

  5. [5]

    Visual Place Recognition: A Survey , journal =

    Lowry, Stephanie and S. Visual Place Recognition: A Survey , journal =. 2016 , volume =

    work page 2016

  6. [6]

    IEEE Access , year =

    Masone, Carlo and Caputo, Barbara , title =. IEEE Access , year =

    work page

  7. [7]

    A Guide to Structureless Visual Localization , journal =

    Panek, Vojtech and Zhou, Qunjie and Ding, Yaqing and Agostinho, S. A Guide to Structureless Visual Localization , journal =. 2026 , volume =

    work page 2026

  8. [8]

    Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) , year =

    Sarlin, Paul-Edouard and Cadena, Cesar and Siegwart, Roland and Dymczyk, Marcin , title =. Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) , year =

    work page

Show all 48 references
  1. [9]

    Long-Term Visual Localization Revisited , journal =

    Toft, Carl and Maddern, Will and Torii, Akihiko and Hammarstrand, Lars and Stenborg, Erik and Safari, Daniel and Okutomi, Masatoshi and Pollefeys, Marc and Sivic, Josef and Pajdla, Tom. Long-Term Visual Localization Revisited , journal =. 2022 , volume =

    2022

  2. [10]

    European Conference on Computer Vision (ECCV) , year =

    Arnold, Eduardo and Wynn, Jamie and Vicente, Sara and Garcia-Hernando, Guillermo and Monszpart, Aron and Prisacariu, Victor and Turmukhambetov, Daniyar and Brachmann, Eric , title =. European Conference on Computer Vision (ECCV) , year =

  3. [11]

    Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) , year =

    Dong, Siyan and Wang, Shuzhe and Liu, Shaohui and Cai, Lulu and Fan, Qingnan and Kannala, Juho and Yang, Yanchao , title =. Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) , year =

  4. [12]

    Structure-from-Motion Revisited , booktitle =

    Sch. Structure-from-Motion Revisited , booktitle =. 2016 , pages =

    2016

  5. [13]

    Proceedings of the IEEE/CVF International Conference on Computer Vision (ICCV) , year =

    Lindenberger, Philipp and Sarlin, Paul-Edouard and Pollefeys, Marc , title =. Proceedings of the IEEE/CVF International Conference on Computer Vision (ICCV) , year =

  6. [14]

    Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition (CVPR) , year =

    Arandjelovi. Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition (CVPR) , year =

  7. [15]

    Fine-Tuning

    Radenovi. Fine-Tuning. IEEE Transactions on Pattern Analysis and Machine Intelligence , year =

  8. [16]

    IEEE Robotics and Automation Letters , year =

    Keetha, Nikhil and Mishra, Avneesh and Karhade, Jay and Jatavallabhula, Krishna Murthy and Scherer, Sebastian and Krishna, Madhava and Garg, Sourav , title =. IEEE Robotics and Automation Letters , year =

  9. [17]

    Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) , year =

    Lu, Feng and Lan, Xiangyuan and Zhang, Lijun and Jiang, Dongmei and Wang, Yaowei and Yuan, Chun , title =. Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) , year =

  10. [18]

    Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) , year =

    Izquierdo, Sergio and Civera, Javier , title =. Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) , year =

  11. [19]

    Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) , year =

    Ali-bey, Amar and Chaib-draa, Brahim and Gigu. Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) , year =

  12. [20]

    Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition Workshops (CVPRW) , year =

    Berton, Gabriele and Masone, Carlo , title =. Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition Workshops (CVPRW) , year =

  13. [21]

    IEEE Robotics and Automation Letters , year =

    Thoma, Janine and Paudel, Danda Pani and Chhatkuli, Ajad and Van Gool, Luc , title =. IEEE Robotics and Automation Letters , year =

  14. [22]

    Benchmarking Image Retrieval for Visual Localization , booktitle =

    Pion, No. Benchmarking Image Retrieval for Visual Localization , booktitle =. 2020 , pages =

    2020

  15. [23]

    Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) , year =

    Wang, Jianyuan and Chen, Minghao and Karaev, Nikita and Vedaldi, Andrea and Rupprecht, Christian and Novotny, David , title =. Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) , year =

  16. [24]

    and Li, Zhenyu and Zhao, Yang and Peng, Sida and Guo, Hengkai and Zhou, Xiaowei and Shi, Guang and Feng, Jiashi and Kang, Bingyi , title =

    Lin, Haotong and Chen, Sili and Liew, Jun Hao and Chen, Donny Y. and Li, Zhenyu and Zhao, Yang and Peng, Sida and Guo, Hengkai and Zhou, Xiaowei and Shi, Guang and Feng, Jiashi and Kang, Bingyi , title =. International Conference on Learning Representations (ICLR) , year =

  17. [25]

    Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) , year =

    Wang, Shuzhe and Leroy, Vincent and Cabon, Yohann and Chidlovskii, Boris and Revaud, J. Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) , year =

  18. [26]

    and Henaff, Mikael and Tang, Hao and Cao, Ang and Chai, Joyce and Meier, Franziska and Feiszli, Matt , title =

    Yang, Jianing and Sax, Alexander and Liang, Kevin J. and Henaff, Mikael and Tang, Hao and Cao, Ang and Chai, Joyce and Meier, Franziska and Feiszli, Matt , title =. Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) , year =

  19. [27]

    International Conference on Learning Representations (ICLR) , year =

    Wang, Yifan and Zhou, Jianjun and Zhu, Haoyi and Chang, Wenzheng and Zhou, Yang and Li, Zizun and Chen, Junyi and Pang, Jiangmiao and Shen, Chunhua and He, Tong , title =. International Conference on Learning Representations (ICLR) , year =

  20. [28]

    and Snavely, Noah and Ho

    Jin, Haian and Wu, Rundi and Zhang, Tianyuan and Gao, Ruiqi and Barron, Jonathan T. and Snavely, Noah and Ho. Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) , year =

  21. [29]

    International Conference on 3D Vision (3DV) , year =

    Keetha, Nikhil Varma and M. International Conference on 3D Vision (3DV) , year =

  22. [30]

    IEEE Transactions on Pattern Analysis and Machine Intelligence , year =

    Umeyama, Shinji , title =. IEEE Transactions on Pattern Analysis and Machine Intelligence , year =

  23. [31]

    , title =

    Rousseeuw, Peter J. , title =. Journal of the American Statistical Association , year =

  24. [32]

    Probabilistic Visual Place Recognition for Hierarchical Localization , journal =

    Xu, Ming and S. Probabilistic Visual Place Recognition for Hierarchical Localization , journal =. 2021 , volume =

    2021

  25. [33]

    Probabilistic Appearance-Invariant Topometric Localization With New Place Awareness , journal =

    Xu, Ming and Fischer, Tobias and S. Probabilistic Appearance-Invariant Topometric Localization With New Place Awareness , journal =. 2021 , volume =

    2021

  26. [34]

    2024 IEEE International Conference on Robotics and Automation (ICRA) , year =

    Suomela, Lauri and Kalliola, Jussi and Edelman, Harry and K. 2024 IEEE International Conference on Robotics and Automation (ICRA) , year =

    2024

  27. [35]

    Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) , year =

    Berton, Gabriele and Masone, Carlo and Caputo, Barbara , title =. Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) , year =

  28. [36]

    Proceedings of the IEEE/CVF Winter Conference on Applications of Computer Vision (WACV) , year =

    Ali-bey, Amar and Chaib-draa, Brahim and Gigu. Proceedings of the IEEE/CVF Winter Conference on Applications of Computer Vision (WACV) , year =

  29. [37]

    Proceedings of the IEEE/CVF International Conference on Computer Vision (ICCV) , year =

    Berton, Gabriele and Trivigno, Gabriele and Caputo, Barbara and Masone, Carlo , title =. Proceedings of the IEEE/CVF International Conference on Computer Vision (ICCV) , year =

  30. [38]

    Proceedings of the IEEE International Conference on Robotics and Automation (ICRA) , year =

    Dellaert, Frank and Fox, Dieter and Burgard, Wolfram and Thrun, Sebastian , title =. Proceedings of the IEEE International Conference on Robotics and Automation (ICRA) , year =

  31. [39]

    The International Journal of Robotics Research , year =

    Cummins, Mark and Newman, Paul , title =. The International Journal of Robotics Research , year =

  32. [40]

    Bayesian Triplet Loss: Uncertainty Quantification in Image Retrieval , booktitle =

    Warburg, Frederik and J. Bayesian Triplet Loss: Uncertainty Quantification in Image Retrieval , booktitle =. 2021 , pages =

    2021

  33. [41]

    Zaffar, Mubariz and Nan, Liangliang and Kooij, Julian F. P. , title =. Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) , year =

  34. [42]

    2022 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS) , year =

    Cai, Kaiwen and Lu, Chris Xiaoxuan and Huang, Xiaowei , title =. 2022 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS) , year =

    2022

  35. [43]

    IEEE Robotics and Automation Letters , year =

    Miller, Emily and Milford, Michael and Hafez, Muhammad Burhan and Ramchurn, Sarvapali and Ehsan, Shoaib , title =. IEEE Robotics and Automation Letters , year =

  36. [44]

    and Wyeth, Gordon F

    Milford, Michael J. and Wyeth, Gordon F. , title =. Proceedings of the IEEE International Conference on Robotics and Automation (ICRA) , year =

  37. [45]

    The International Journal of Robotics Research , year =

    Pronobis, Andrzej and Caputo, Barbara , title =. The International Journal of Robotics Research , year =

  38. [46]

    Proceedings of the German Conference on Pattern Recognition (GCPR) , series =

    Wenzel, Patrick and Wang, Rui and Yang, Nan and Cheng, Qing and Khan, Qadeer and von Stumberg, Lukas and Zeller, Niclas and Cremers, Daniel , title =. Proceedings of the German Conference on Pattern Recognition (GCPR) , series =. 2021 , pages =

    2021

  39. [47]

    The International Journal of Robotics Research , year =

    Maddern, Will and Pascoe, Geoffrey and Linegar, Chris and Newman, Paul , title =. The International Journal of Robotics Research , year =

  40. [48]

    arXiv preprint arXiv:2002.10152 , year =

    Maddern, Will and Pascoe, Geoffrey and Gadd, Matthew and Barnes, Dan and Yeomans, Brian and Newman, Paul , title =. arXiv preprint arXiv:2002.10152 , year =

    2002 arXiv

Pith tools

Reviewed August 7, 2026 · model on record in the stance chip above.