Pith. sign in

REVIEW 3 major objections 6 minor 55 references

SuperPlace: The Renaissance of Classical Feature Aggregation for Visual Place Recognition in the Era of Foundation Models

T0 review · 3 major / 6 minor · reviewed 2026-08-07 · deepseek-v4-flash

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

Pith's one-line read Lightly upgraded GeM and NetVLAD aggregators, trained on multiple datasets with aligned labels, match or beat recent VPR methods at one-tenth the descriptor size, and one variant ranks first on MSLS.

desk verdict A solid empirical paper with a real efficiency contribution, but the G2M accuracy gain is partly selected on a test set that is then reported as final, and the headline SOTA comparison is confounded by training-set differences. read the letter →

arxiv 2506.13073 v1 pith:RUNLRFU5 submitted 2025-06-16 cs.CV

Bingxi Liu , Pengju Zhang , Li He , Hao Chen , Shiyi Guo , Yihong Wu , Jinqiang Cui , Hong Zhang This is my paper · ORCID
classification cs.CV
keywords visualplacerecognitionfeatureaggregationGeMpoolingNetVLADDINOv2multi-datasettraininglabelalignmentcompactdescriptors
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

The paper's thesis is that the recently invented feature-aggregation modules in visual place recognition are not necessary: the classical GeM and NetVLAD pooling modules, upgraded in small ways and trained on multiple datasets at once, match or exceed them with far smaller descriptors. The upgrades are a second GeM that learns the principal components of a feature map across channels and uses them to calibrate the first GeM's output, and a two-stage fine-tuning schedule in which NetVLAD first learns a high-dimensional representation and then compresses it through a single linear layer. A third ingredient, supervised label alignment, turns GPS-distance labels into class labels using UTM grid cells and local feature matching, so four large datasets can be pooled into one training framework. If the claims hold, a 768-dimensional descriptor is enough for competitive place recognition, and the remaining large gains come from training on more data rather than from inventing more complex aggregators.

What carries the argument

Three mechanisms carry the argument. Generalized Channel Attention (GCA) is a second GeM pooling branch that computes a channel-wise summary, passes it through a low-rank MLP, GELU, and sigmoid, and re-weights the main GeM's descriptor; it is the component that lets the compact model stay at 768 dimensions. Secondary fine-tuning (FT2) applies to NetVLAD-Linear: first the backbone and NetVLAD are fine-tuned in a high-dimensional space, then the single linear compression layer is fine-tuned while the rest is frozen; this two-stage schedule is what makes the linear layer competitive with PCA-based compression. Supervised label alignment (SLA) converts the distance-based supervision of four datasets into class labels by partitioning UTM coordinates into grid cells, adding heading classes where available, and refining classes with local feature matching, so one model can train on all datasets simultaneously.

What would settle it

Retrain G2M with the GCA rank and activation selected on a validation split that is disjoint from every reported test set, then evaluate on Pitts-30k-test; if the Recall@1 advantage over plain GeM shrinks to well below the reported +0.7, the compact-aggregation claim is partly an artifact of test-set tuning.

Watch

Extended reading notes

Core claim

The central discovery is that heavily engineered aggregation layers are not the source of recent progress in VPR. Starting from DINOv2 as a pre-trained backbone, the paper shows that a plain GeM with the last four layers fine-tuned already achieves strong results, and that adding a Generalized Channel Attention (GCA) branch — a second GeM pooling that summarizes each channel, passes it through a low-rank MLP with GELU and sigmoid, and re-weights the output of the first GeM — improves accuracy while keeping the descriptor at 768 dimensions. For high-dimensional retrieval, the paper shows that a linear projection after NetVLAD underperforms until it is trained in a second stage after the high-dimensional NetVLAD representation is already learned; this secondary fine-tuning updates only 0.11% of the model's parameters and closes most of the gap with full NetVLAD. The supervised label alignment then combines GSV-Cities, Pittsburgh-250k, MSLS, and SF-XL into one training mixture, and this multi-dataset training is what pushes the NetVLAD variant to first place on the MSLS challenge leaderboard and gives the 768-dimensional model its strongest results.

Load-bearing premise

The load-bearing premise is that the gain of the new two-GeM module over a plain GeM was not inflated by tuning its rank and activation on the same Pittsburgh test set that is later reported as the final result; the paper notes in Section 4.4 that those settings were selected using Pitts-30k-test outcomes.

Editorial extensions

If this is right

  • SP-G2M's 768-dimensional descriptors imply that large VPR databases can be searched with roughly one-tenth the memory and computation of recent methods that use 4,096 to 12,288 dimensions.
  • NVL-FT2 reaches first place on the MSLS challenge leaderboard with a single-stage pipeline, so two-stage re-ranking is not required for top-level accuracy.
  • Training on the aligned mixture helps on every test set that matches one of the constituent datasets, indicating that the four datasets carry complementary information rather than being redundant.
  • The FT2 second stage is cheap because it updates only 0.11% of the model's parameters, making the accuracy gain over one-shot NetVLAD-Linear available at low extra training cost.
  • With DINOv2-L and the aligned training mixture, SP-NVL-FT2 reaches 84.8% Recall@1 on MSLS-challenge, which the paper reports as 11.3 points above a strong re-ranking baseline.

Reading between the lines

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

  • An extension the paper does not run: attach the same GCA calibration branch to other aggregators such as SALAD or BoQ and see whether the 768-dimension advantage persists; if it does, channel calibration is a general cheap upgrade rather than a GeM-specific fix.
  • The SLA recipe generalizes beyond VPR: any geo-tagged image task with inconsistent supervision could reuse the UTM-grid-plus-local-matching trick to merge datasets, turning multi-source metric learning into a classification problem.
  • If the 768-dimension result survives a strict tuning-free evaluation, the practical bottleneck for large-scale place recognition shifts from descriptor expressiveness to database search cost, and re-ranking becomes optional rather than necessary for top performance.
Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

3 major / 6 minor

Summary. The paper proposes SuperPlace, a VPR framework built on DINOv2, with two classical aggregators: G2M (GeM plus a channel-attention branch) and NVL-FT2 (NetVLAD with a linear compression layer trained with a second fine-tuning stage), together with a supervised label alignment (SLA) method that unifies training labels across GSV-Cities, Pitts-250k, MSLS, and SF-XL. Experiments compare SuperPlace against recent VPR methods on 12 test sets, report ablations of each component, and claim that G2M reaches competitive accuracy with 768-d descriptors while NVL-FT2 ranks first on the MSLS leaderboard.

Significance. The paper's strength is its systematic ablation methodology: within each table, methods are trained on the same set, isolating the effect of the aggregator (Tables 7, 10, 11) and of the training-set composition (Table 5). If the claims held, the paper would show that simple classical aggregators remain competitive with more complex recent designs at a fraction of the descriptor dimensionality, and that multi-dataset training via label alignment is a valuable recipe. Efficiency figures (parameter counts, inference times) are also reported. However, the headline SOTA comparison is confounded by training-set differences, and the GCA hyperparameters are selected on a test set that is later reported as a benchmark, which undermines the controlled evidence for G2M's advantage.

major comments (3)
  1. [Section 4.4, Table 8] The paper explicitly states that GCA rank and activation were selected 'mainly ... based on the results of Pitts-30k' (Section 4.4), and the same Pitts-30k-test split is used as a final benchmark in Tables 3, 7, 10, and 11. In Table 8, rank=64 with GELU improves over GeM on Pitts-30k-test (+0.7 R@1) but is worse on Tokyo-24/7 (-0.3) and only marginally better on MSLS-val (+0.1); on Tokyo, rank=3 and ReLU are both better than the chosen configuration. In Table 10 (GPMS training), G2M is actually 0.2 R@1 below GeM on Pitts-30k-test. This constitutes selection on the test set, and the small, inconsistent gains mean the 'one-tenth dimensions' claim is not independently established. Please validate the GCA configuration on a held-out validation set (e.g., a separated split or a different benchmark), report multiple seeds or confidence intervals, and avoid using the benchmark set for model selection.
  2. [Table 3 and Abstract] The abstract's statement that 'G2M achieves promising results with only one-tenth of the feature dimensions compared to recent methods' is based primarily on Table 3, where SP-G2M and SP-NVL-FT2 are trained on GPMS while SALAD, BoQ, and other baselines use GSV-Cities (this is acknowledged in Section 4.1). The different training sets make the comparison non-matching; the performance gap conflates the aggregation design with the SLA contribution. Please provide a matched comparison where baselines are also trained on GPMS (or SuperPlace on GSV-Cities only), or explicitly separate the contribution of SLA from the aggregation gains in the claims.
  3. [Table 10] Under GPMS training, the GCA module actually hurts performance on Pitts-30k-test (92.2 vs 92.4 for GeM) and leaves G2M below SALAD (92.8). This result is inconsistent with the claim that G2M is a generally superior low-dimensional aggregation. Please temper the claim or provide an analysis of why the benefit disappears with the multi-dataset training recipe.
minor comments (6)
  1. [Section 2] The sentence 'The use of VFMs in VPR has been limited compared to other vision tasks, compared to other vision tasks' repeats 'compared to other vision tasks'.
  2. [Figure 3 caption] The phrase 'between the patches' should likely be 'on the patches' or 'across the patches'.
  3. [Table 6] The training set for SALAD-SLA is not specified; please state it to make the comparison with SALAD-CM controlled.
  4. [Section 3.3] The number of images after label alignment is said to be 'recorded in Tab. 2', but Table 2 lists test sets; please add the aligned training-set statistics.
  5. [Table 9] The row 'NVL-MLP✓8192' is ambiguous about whether the CLS token is concatenated; please define the checkmark columns clearly.
  6. [Abstract and Section 1] The claim that NVL-FT2 'ranks first on the MSLS leaderboard' is not verifiable from the manuscript; please provide a citation, link, or snapshot with a date.

Circularity Check

1 steps flagged · score 5.0 of 10

G2M's controlled accuracy gain over GeM is selected on the same Pitts-30k-test split later reported as the final benchmark, so the compact-aggregation evidence is partly fitted; other claims remain independent.

  1. fitted input called prediction [Section 4.4 (Ablation studies of the improved GeM), Table 8; see also Section 4.1 and Tables 3 and 7]
    "Since the distributions of GSV-Cities and Pitts-30k were closely related, we mainly selected parameters based on the results of Pitts-30k."

    Table 8 reports Pitts-30k-test, Tokyo-24/7, and MSLS-val R@1 for GCA ranks 3, 32, 64, 128 and GELU/ReLU activations. Rank 64 with GELU is best only on Pitts-30k-test (92.6), while on Tokyo-24/7 both rank 3 (95.2) and ReLU (94.9) beat it, and on MSLS-val ranks 3, 32, and 128 all beat it. Section 4.1 then fixes the model as 'rank of 64, GELU' based on this table. The same Pitts-30k-test split is subsequently reported as a final benchmark in Tables 3 and 7, and the controlled comparison 'GeM + GCA (G2M)' at 92.6 versus 'FT4-GeM (Our impl.)' at 91.9 is exactly the selection maximum from Table 8. Therefore the +0.7 R@1 margin claimed for GCA over plain GeM is fitted to the validation split, not an independent confirmation of the contribution.

full rationale

This is an empirical paper with no formal derivation chain, so the classic self-definitional and self-citation circularities do not apply. No load-bearing self-citation is present: the cited prior work (GeM, NetVLAD, SALAD, BoQ, DINOv2, etc.) is external to the authors, and no uniqueness theorem is imported from the authors' own prior work. The identified circularity is a selection loop: Section 4.4 explicitly states that GCA hyperparameters were mainly chosen using Pitts-30k-test results, and Section 4.1 hard-codes rank 64 and GELU. The same Pitts-30k-test split is then reported as a final benchmark in Tables 3, 7, 10, and 11, and the controlled GCA gain is the Table 8 maximum on that split. On the other two benchmarks in Table 8 the chosen configuration is not best, and under GPMS training (Table 10) G2M is 0.2 points below plain GeM on Pitts-30k-test, so the magnitude and even the sign of GCA's benefit over GeM is not robust outside the selection split. However, the one-tenth-dimension efficiency claim is fixed by architecture (768-dimensional output), the MSLS-challenge leaderboard result for NVL-FT2 is a held-out online evaluation, and the competitiveness of plain DINOv2-GeM is independent of this tuning. These independent supports prevent the paper from being wholly circular, but the controlled validation of the headline GCA contribution is compromised, warranting a mid-range score.

Assumptions & free parameters 9 free parameters · 5 assumptions · 0 invented entities

The central claims depend on a handful of hand-set hyperparameters (GCA rank and activation, descriptor dimensions, NetVLAD K, SLA grid parameters), several of which were selected using the Pitts-30k test set. The work also relies on domain assumptions about label alignment and the suitability of DINOv2 features. No new physical or conceptual entities are introduced.

free parameters (9)
  • GCA rank r = 64
    Selected in Table 8 based on Pitts-30k-test results; controls the low-rank MLP capacity and affects G2M accuracy.
  • GCA activation function = GELU
    Chosen via ablation against ReLU in Table 8; slightly better on Tokyo 24/7 and comparable elsewhere.
  • G2M descriptor dimension = 768
    Matches the DINOv2-B channel count; central to the compactness claim of one-tenth the dimensions of recent methods.
  • NetVLAD cluster count K = 64
    Taken from NetVLAD prior work, but it directly sets the NVL high-dimensional space and the final 8192-dim descriptor.
  • NVL output dimension = 8192
    Design choice balancing accuracy and speed; used in all NVL-FT2 experiments.
  • SLA grid cell size M = 10 m
    Inherited from CosPlace; defines the UTM class granularity in Eq. (2).
  • SLA heading bin alpha = 30 deg
    Inherited from CosPlace; used for SF-XL classes, not for Pittsburgh or MSLS.
  • CosPlace group strategy N, L = 5, 2
    Used to overcome quantization errors in SF-XL labels, following CosPlace.
  • Training resolution = 322x322 for main tables, 224x224 for ablations
    Higher resolution improves accuracy; the choice affects the reported numbers in Tables 3-4 versus Tables 5-11.
assumptions (5)
  • domain assumption UTM grid cells of 10 m and 30-degree heading bins produce class labels that are meaningful for VPR training.
    Invoked in Sec. 3.3, Eq. (2), following CosPlace's setup for SF-XL; all SLA label construction relies on this.
  • domain assumption Local feature matching (LightGlue or SuperGlue) can correctly align panoramic sub-images to the same UTM grid cell in Pittsburgh-250k and MSLS.
    Stated in Sec. 3.3; incorrect matches would corrupt the P and M class labels and hurt training.
  • domain assumption Multi-similarity loss is an appropriate objective for learning VPR embeddings.
    Adopted from prior work [49] and used for all SuperPlace training runs.
  • domain assumption DINOv2 patch tokens, reshaped into a C x H x W feature map, provide a representation whose channel statistics are useful for place recognition.
    The whole G2M design (GeM plus channel attention) assumes this representation; motivated by the PCA visualizations in Fig. 3.
  • standard math GeM pooling, softmax assignment, and L2 normalization are standard building blocks with their usual mathematical properties.
    Used in Eq. (1) and in the NetVLAD formulation without re-derivation.

how reviews work

0 comments
Cite this review

Pith. "Pith review of SuperPlace: The Renaissance of Classical Feature Aggregation for Visual Place Recognition in the Era of Foundation Models." pith.science (2026). https://pith.science/paper/RUNLRFU5

@misc{pith2026250613073,
  author       = {Pith},
  title        = {Pith review of: SuperPlace: The Renaissance of Classical Feature Aggregation for Visual Place Recognition in the Era of Foundation Models},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/RUNLRFU5}},
  note         = {Machine review of arXiv:2506.13073}
}
abstract

Recent visual place recognition (VPR) approaches have leveraged foundation models (FM) and introduced novel aggregation techniques. However, these methods have failed to fully exploit key concepts of FM, such as the effective utilization of extensive training sets, and they have overlooked the potential of classical aggregation methods, such as GeM and NetVLAD. Building on these insights, we revive classical feature aggregation methods and develop more fundamental VPR models, collectively termed SuperPlace. First, we introduce a supervised label alignment method that enables training across various VPR datasets within a unified framework. Second, we propose G$^2$M, a compact feature aggregation method utilizing two GeMs, where one GeM learns the principal components of feature maps along the channel dimension and calibrates the output of the other. Third, we propose the secondary fine-tuning (FT$^2$) strategy for NetVLAD-Linear (NVL). NetVLAD first learns feature vectors in a high-dimensional space and then compresses them into a lower-dimensional space via a single linear layer. Extensive experiments highlight our contributions and demonstrate the superiority of SuperPlace. Specifically, G$^2$M achieves promising results with only one-tenth of the feature dimensions compared to recent methods. Moreover, NVL-FT$^2$ ranks first on the MSLS leaderboard.

Figures

Figures reproduced from arXiv: 2506.13073 by the authors.

Figure 1
Figure 1. The Recall@1 and descriptor dimensionality compari [PITH_FULL_IMAGE:figures/full_fig_p001_1.png] view at source ↗
Figure 2
Figure 2. Illustration of two improved classical aggregations. Without all the bells and whistles, we improved on classic aggregations by adding simple structures, making them better than recent complex aggregation layers with many parameters. learnable global queries to probe the input features through cross-attention, ensuring consistent information aggrega￾tion. SALAD [23] redefined the soft assignment of local features in… view at source ↗
Figure 3
Figure 3. Visualization of feature maps weighted by different components. We computed a PCA between the patches of the images from the AmsterTime dataset and showed their first three components. We found that high and low response areas of feature maps after principal component weighting strongly correlate with the VPR task. a high-dimensional space, and (2) estimating an unsuper￾vised model for high-to-low dimensional projec… view at source ↗
Figures from the paper (1 more)
Figure 4
Figure 4. Figure 4: Schematic diagram of collecting VPR data. VPR im￾ages with the same label are drawn using the same color in each minimal grid map. Although orange triangles appear in all four subgraphs, they represent different labels in each. The black trian￾gles indicate that images…

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

55 extracted references · 37 canonical work pages

  1. [1]

    Gsv-cities: Toward appropriate supervised visual place recognition.Neurocomputing, 513:194–203, 2022

    Amar Ali-bey, Brahim Chaib-draa, and Philippe Gigu `ere. Gsv-cities: Toward appropriate supervised visual place recognition.Neurocomputing, 513:194–203, 2022. 1, 2, 4

    work page 2022

  2. [2]

    Mixvpr: Feature mixing for visual place recognition

    Amar Ali-bey, Brahim Chaib-draa, and Philippe Gigu `ere. Mixvpr: Feature mixing for visual place recognition. InPro- ceedings of the IEEE/CVF Winter Conference on Applica- tions of Computer Vision, pages 2998–3007, 2023. 3, 6

    work page 2023

  3. [3]

    Boq: A place is worth a bag of learnable queries

    Amar Ali-bey, Brahim Chaib-draa, and Philippe Gigu `ere. Boq: A place is worth a bag of learnable queries. InPro- ceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pages 17794–17803, 2024. 2, 3, 6

    work page 2024

  4. [4]

    Fast and incremental method for loop-closure detection using bags of visual words.IEEE transactions on robotics, 24(5):1027–1037, 2008

    Adrien Angeli, David Filliat, St ´ephane Doncieux, and Jean- Arcady Meyer. Fast and incremental method for loop-closure detection using bags of visual words.IEEE transactions on robotics, 24(5):1027–1037, 2008. 1, 2

    2008

  5. [5]

    Google street view: Capturing the world at street level.Computer, 43:32–38, 2010

    Dragomir Anguelov, Carole Dulong, Daniel Filip, Christian Frueh, St ´ephane Lafon, Richard Lyon, Abhijit Ogale, Luc Vincent, and Josh Weaver. Google street view: Capturing the world at street level.Computer, 43:32–38, 2010. 2

    work page 2010

  6. [6]

    Netvlad: Cnn architecture for weakly supervised place recognition

    Relja Arandjelovic, Petr Gronat, Akihiko Torii, Tomas Pa- jdla, and Josef Sivic. Netvlad: Cnn architecture for weakly supervised place recognition. InProceedings of the IEEE Conference on Computer Vision and Pattern Recognition, pages 5297–5307, 2016. 1, 2, 3, 4, 5, 6, 8

    work page 2016

  7. [7]

    Speeded-up robust features (surf).Computer vi- sion and image understanding, 110(3):346–359, 2008

    Herbert Bay, Andreas Ess, Tinne Tuytelaars, and Luc Van Gool. Speeded-up robust features (surf).Computer vi- sion and image understanding, 110(3):346–359, 2008. 1, 2

    work page 2008

  8. [8]

    Viewpoint invariant dense matching for vi- sual geolocalization

    Gabriele Berton, Carlo Masone, Valerio Paolicelli, and Bar- bara Caputo. Viewpoint invariant dense matching for vi- sual geolocalization. InProceedings of the IEEE/CVF In- ternational Conference on Computer Vision (ICCV), pages 12169–12178, 2021. 1

    work page 2021

Show all 55 references
  1. [9]

    Re- thinking visual geo-localization for large-scale applications

    Gabriele Berton, Carlo Masone, and Barbara Caputo. Re- thinking visual geo-localization for large-scale applications. InProceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pages 4878–4888, 2022. 1, 2, 3, 4, 5, 6

    2022

  2. [10]

    Deep visual geo-localization benchmark

    Gabriele Berton, Riccardo Mereu, Gabriele Trivigno, Carlo Masone, Gabriela Csurka, Torsten Sattler, and Barbara Ca- puto. Deep visual geo-localization benchmark. InProceed- ings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), pages 5396–5407, 2022. 1, 6

    2022

  3. [11]

    Eigenplaces: Training viewpoint robust models for visual place recognition

    Gabriele Berton, Gabriele Trivigno, Barbara Caputo, and Carlo Masone. Eigenplaces: Training viewpoint robust models for visual place recognition. InProceedings of the IEEE/CVF International Conference on Computer Vision, pages 11080–11090, 2023. 3, 6

    2023

  4. [12]

    Deep learning features at scale for visual place recognition

    Zetao Chen, Adam Jacobson, Niko S ¨underhauf, Ben Up- croft, Lingqiao Liu, Chunhua Shen, Ian Reid, and Michael Milford. Deep learning features at scale for visual place recognition. In2017 IEEE international conference on robotics and automation (ICRA), pages 3223–3230. IEEE,

  5. [13]

    Fisher vectors: Be- yond bag-of-visual-words image representations

    Gabriela Csurka and Florent Perronnin. Fisher vectors: Be- yond bag-of-visual-words image representations. InInterna- tional conference on computer vision, imaging and computer graphics, pages 28–42. Springer, 2010. 2

    2010

  6. [14]

    Learning to reduce scale differences for large- scale invariant image matching.IEEE Transactions on Cir- cuits and Systems for Video Technology, 2022

    Yujie Fu, Pengju Zhang, Bingxi Liu, Zheng Rong, and Yi- hong Wu. Learning to reduce scale differences for large- scale invariant image matching.IEEE Transactions on Cir- cuits and Systems for Video Technology, 2022. 1

    2022

  7. [15]

    Self-supervising fine-grained region similarities for large-scale image localization

    Yixiao Ge, Haibo Wang, Feng Zhu, Rui Zhao, and Hong- sheng Li. Self-supervising fine-grained region similarities for large-scale image localization. InComputer Vision– ECCV 2020: 16th European Conference, Glasgow, UK, Au- gust 23–28, 2020, Proceedings, Part IV 16, pages 369–38...

    2020

  8. [16]

    Ghost-dil-netvlad: a lightweight neural network for visual place recognition.arXiv preprint arXiv:2112.11679, 2021

    Qingyuan Gong, Yu Liu, Liqiang Zhang, and Renhe Liu. Ghost-dil-netvlad: a lightweight neural network for visual place recognition.arXiv preprint arXiv:2112.11679, 2021. 2

    2021 arXiv

  9. [17]

    Design space exploration of low-bit quantized neural networks for visual place recogni- tion.IEEE Robotics and Automation Letters, 2024

    Oliver Grainge, Michael Milford, Indu Bodala, Sarvapali D Ramchurn, and Shoaib Ehsan. Design space exploration of low-bit quantized neural networks for visual place recogni- tion.IEEE Robotics and Automation Letters, 2024. 3

    2024

  10. [18]

    Patch-netvlad: Multi-scale fusion of locally-global descriptors for place recognition

    Stephen Hausler, Sourav Garg, Ming Xu, Michael Milford, and Tobias Fischer. Patch-netvlad: Multi-scale fusion of locally-global descriptors for place recognition. InProceed- ings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pages 14141–14152, 2021. 6

    2021

  11. [19]

    Distill- ing the knowledge in a neural network.arXiv preprint arXiv:1503.02531, 2015

    Geoffrey Hinton, Oriol Vinyals, and Jeff Dean. Distill- ing the knowledge in a neural network.arXiv preprint arXiv:1503.02531, 2015. 2

    2015 arXiv

  12. [20]

    Bocnf: efficient im- age matching with bag of convnet features for scalable and robust visual place recognition.Autonomous Robots, page 1169–1185, 2018

    Yi Hou, Hong Zhang, and Shilin Zhou. Bocnf: efficient im- age matching with bag of convnet features for scalable and robust visual place recognition.Autonomous Robots, page 1169–1185, 2018. 2

    2018

  13. [21]

    Lora: Low-rank adaptation of large language models.arXiv preprint arXiv:2106.09685, 2021

    Edward J Hu, Yelong Shen, Phillip Wallis, Zeyuan Allen- Zhu, Yuanzhi Li, Shean Wang, Lu Wang, and Weizhu Chen. Lora: Low-rank adaptation of large language models.arXiv preprint arXiv:2106.09685, 2021. 4

    2021 arXiv

  14. [22]

    Squeeze-and-excitation net- works

    Jie Hu, Li Shen, and Gang Sun. Squeeze-and-excitation net- works. InProceedings of the IEEE conference on computer vision and pattern recognition, pages 7132–7141, 2018. 4, 8

    2018

  15. [23]

    Optimal transport ag- gregation for visual place recognition

    Sergio Izquierdo and Javier Civera. Optimal transport ag- gregation for visual place recognition. InProceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), 2024. 2, 3, 4, 5, 6, 7, 8

    2024

  16. [24]

    Close, but not there: Boosting geographic distance sensitivity in visual place recognition.arXiv preprint arXiv:2407.02422, 2024

    Sergio Izquierdo and Javier Civera. Close, but not there: Boosting geographic distance sensitivity in visual place recognition.arXiv preprint arXiv:2407.02422, 2024. 2, 6, 7

    2024 arXiv

  17. [25]

    Aggregating local descriptors into a compact image representation

    Herv ´e J´egou, Matthijs Douze, Cordelia Schmid, and Patrick P´erez. Aggregating local descriptors into a compact image representation. In2010 IEEE computer society conference on computer vision and pattern recognition, pages 3304–

  18. [26]

    Anyloc: Towards universal vi- sual place recognition.IEEE Robotics and Automation Let- ters, 2023

    Nikhil Keetha, Avneesh Mishra, Jay Karhade, Kr- ishna Murthy Jatavallabhula, Sebastian Scherer, Madhava Krishna, and Sourav Garg. Anyloc: Towards universal vi- sual place recognition.IEEE Robotics and Automation Let- ters, 2023. 2

    2023

  19. [27]

    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 Con- ference on Computer Vision, pages 4015–4026, 2023. 2, 3

    2023

  20. [28]

    LightGlue: Local Feature Matching at Light Speed

    Philipp Lindenberger, Paul-Edouard Sarlin, and Marc Polle- feys. LightGlue: Local Feature Matching at Light Speed. In ICCV, 2023. 5

    2023

  21. [29]

    A flexible and efficient loop closure detection based on motion knowledge

    Bingxi Liu, Fulin Tang, Yujie Fu, Yanqun Yang, and Yihong Wu. A flexible and efficient loop closure detection based on motion knowledge. In2021 IEEE International Conference on Robotics and Automation (ICRA), pages 11241–11247. IEEE, 2021. 1

    2021

  22. [30]

    Npr: Nocturnal place recognition us- ing nighttime translation in large-scale training procedures

    Bingxi Liu, Yujie Fu, Feng Lu, Jinqiang Cui, Yihong Wu, and Hong Zhang. Npr: Nocturnal place recognition us- ing nighttime translation in large-scale training procedures. IEEE Journal of Selected Topics in Signal Processing, 18 (3):368–379, 2024. 1

    2024

  23. [31]

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

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

    2004

  24. [32]

    Visual place recognition: A survey.ieee transactions on robotics, 32:1–19, 2015

    Stephanie Lowry, Niko S ¨underhauf, Paul Newman, John J Leonard, David Cox, Peter Corke, and Michael J Milford. Visual place recognition: A survey.ieee transactions on robotics, 32:1–19, 2015. 1

    2015

  25. [33]

    Cricavpr: Cross-image correlation-aware representation learning for visual place recognition

    Feng Lu, Xiangyuan Lan, Lijun Zhang, Dongmei Jiang, Yaowei Wang, and Chun Yuan. Cricavpr: Cross-image correlation-aware representation learning for visual place recognition. InProceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pages 16772– 16782...

    2024

  26. [34]

    Towards seamless adapta- tion of pre-trained models for visual place recognition.arXiv preprint arXiv:2402.14505, 2024

    Feng Lu, Lijun Zhang, Xiangyuan Lan, Shuting Dong, Yaowei Wang, and Chun Yuan. Towards seamless adapta- tion of pre-trained models for visual place recognition.arXiv preprint arXiv:2402.14505, 2024. 2, 3, 6

    2024 arXiv

  27. [35]

    Seqslam: Visual route-based navigation for sunny summer days and stormy winter nights

    Michael J Milford and Gordon F Wyeth. Seqslam: Visual route-based navigation for sunny summer days and stormy winter nights. In2012 IEEE international conference on robotics and automation, pages 1643–1649. IEEE, 2012. 2

    2012

  28. [36]

    Dinov2: Learning robust visual features without super- vision.Transactions on Machine Learning Research, 2023

    Maxime Oquab, Timoth ´ee Darcet, Th´eo Moutakanni, Huy V V o, Marc Szafraniec, Vasil Khalidov, Pierre Fernandez, Daniel HAZIZA, Francisco Massa, Alaaeldin El-Nouby, et al. Dinov2: Learning robust visual features without super- vision.Transactions on Machine Learning Research, 2023. 2

    2023

  29. [37]

    At- tentional pyramid pooling of salient visual residuals for place recognition

    Guohao Peng, Jun Zhang, Heshan Li, and Danwei Wang. At- tentional pyramid pooling of salient visual residuals for place recognition. In2021 IEEE/CVF International Conference on Computer Vision (ICCV), 2021. 2

    2021

  30. [38]

    Fine- tuning cnn image retrieval with no human annotation.IEEE transactions on pattern analysis and machine intelligence, 41(7):1655–1668, 2018

    Filip Radenovi ´c, Giorgos Tolias, and Ond ˇrej Chum. Fine- tuning cnn image retrieval with no human annotation.IEEE transactions on pattern analysis and machine intelligence, 41(7):1655–1668, 2018. 2, 3

    2018

  31. [39]

    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. In2011 International conference on computer vision, pages 2564–

  32. [40]

    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. 5

    2020

  33. [41]

    Benchmarking 6dof outdoor visual localization in changing conditions

    Torsten Sattler, Will Maddern, Carl Toft, Akihiko Torii, Lars Hammarstrand, Erik Stenborg, Daniel Safari, Masatoshi Okutomi, Marc Pollefeys, Josef Sivic, et al. Benchmarking 6dof outdoor visual localization in changing conditions. In Proceedings of the IEEE conference on compu...

    2018

  34. [42]

    Are we there yet? challenging seqslam on a 3000 km journey across all four seasons

    Niko S ¨underhauf, Peer Neubert, and Peter Protzel. Are we there yet? challenging seqslam on a 3000 km journey across all four seasons. InProc. of workshop on long-term auton- omy, IEEE international conference on robotics and automa- tion (ICRA), page 2013, 2013. 6

    2013

  35. [43]

    Long-term visual localization revisited.IEEE Transactions on Pattern Analy- sis and Machine Intelligence, 44:2074–2088, 2020

    Carl Toft, Will Maddern, Akihiko Torii, Lars Hammarstrand, Erik Stenborg, Daniel Safari, Masatoshi Okutomi, Marc Pollefeys, Josef Sivic, Tomas Pajdla, et al. Long-term visual localization revisited.IEEE Transactions on Pattern Analy- sis and Machine Intelligence, 44:2074–2088, 2020. 1

    2020

  36. [44]

    Visual place recognition with repetitive structures

    Akihiko Torii, Josef Sivic, Masatoshi Okutomi, and Tomas Pajdla. Visual place recognition with repetitive structures. IEEE Transactions on Pattern Analysis and Machine Intelli- gence, 37:2346–2359, 2015. 1

    2015

  37. [45]

    24/7 place recognition by view synthesis.IEEE Transactions on Pattern Analysis and Ma- chine Intelligence, page 14, 2017

    Akihiko Torii, Relja Arandjelovic, Josef Sivic, Masatoshi Okutomi, and Tomas Pajdla. 24/7 place recognition by view synthesis.IEEE Transactions on Pattern Analysis and Ma- chine Intelligence, page 14, 2017. 5

    2017

  38. [46]

    Ef- fovpr: Effective foundation model utilization for visual place recognition.arXiv preprint arXiv:2405.18065, 2024

    Issar Tzachor, Boaz Lerner, Matan Levy, Michael Green, Tal Berkovitz Shalev, Gavriel Habib, Dvir Samuel, Noam Korngut Zailer, Or Shimshi, Nir Darshan, et al. Ef- fovpr: Effective foundation model utilization for visual place recognition.arXiv preprint arXiv:2405.18065, 2024. 6

    2024

  39. [47]

    Transvpr: Transformer-based place recognition with multi-level attention aggregation

    Ruotong Wang, Yanqing Shen, Weiliang Zuo, Sanping Zhou, and Nanning Zheng. Transvpr: Transformer-based place recognition with multi-level attention aggregation. InPro- ceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pages 13648–13657, 2022. 6

    2022

  40. [48]

    Dust3r: Geometric 3d vi- sion made easy

    Shuzhe Wang, Vincent Leroy, Yohann Cabon, Boris Chidlovskii, and Jerome Revaud. Dust3r: Geometric 3d vi- sion made easy. InCVPR, 2024. 2, 3

    2024

  41. [49]

    Multi-similarity loss with general pair weighting for deep metric learning

    Xun Wang, Xintong Han, Weilin Huang, Dengke Dong, and Matthew R Scott. Multi-similarity loss with general pair weighting for deep metric learning. InProceedings of the IEEE Conference on Computer Vision and Pattern Recogni- tion, pages 5022–5030, 2019. 3, 4

    2019

  42. [50]

    Mapillary street-level sequences: A dataset for lifelong place recognition

    Frederik Warburg, Soren Hauberg, Manuel Lopez- Antequera, Pau Gargallo, Yubin Kuang, and Javier Civera. Mapillary street-level sequences: A dataset for lifelong place recognition. InProceedings of the IEEE/CVF conference on computer vision and pattern recognition, pages 2626–2...

    2020

  43. [51]

    Cbam: Convolutional block attention module

    Sanghyun Woo, Jongchan Park, Joon-Young Lee, and In So Kweon. Cbam: Convolutional block attention module. In Proceedings of the European conference on computer vision (ECCV), pages 3–19, 2018. 8

    2018

  44. [52]

    Asymmetric feature fusion for image retrieval

    Hui Wu, Min Wang, Wengang Zhou, Zhenbo Lu, and Houqiang Li. Asymmetric feature fusion for image retrieval. InProceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pages 11082–11092, 2023. 3

    2023

  45. [53]

    Depth anything: Unleashing the power of large-scale unlabeled data

    Lihe Yang, Bingyi Kang, Zilong Huang, Xiaogang Xu, Jiashi Feng, and Hengshuang Zhao. Depth anything: Unleashing the power of large-scale unlabeled data. InProceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pages 10371–10381, 2024. 2, 3

    2024

  46. [54]

    Amstertime: A visual place recognition benchmark dataset for severe domain shift

    Burak Yildiz, Seyran Khademi, Ronald Maria Siebes, and Jan Van Gemert. Amstertime: A visual place recognition benchmark dataset for severe domain shift. In2022 26th International Conference on Pattern Recognition (ICPR), pages 2749–2755. IEEE, 2022. 6

    2022

  47. [55]

    R2former: Unified retrieval and reranking transformer for place recognition

    Sijie Zhu, Linjie Yang, Chen Chen, Mubarak Shah, Xiao- hui Shen, and Heng Wang. R2former: Unified retrieval and reranking transformer for place recognition. InProceedings of the IEEE/CVF Conference on Computer Vision and Pat- tern Recognition, pages 19370–19380, 2023. 6 11

    2023

Pith tools

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