Pith. sign in

REVIEW 3 major objections 4 minor 32 references

Cross-View Sequential Visual Localization with Spatio-Temporal Context Modeling for Autonomous Driving

T0 review · 3 major / 4 minor · reviewed 2026-08-12 · deepseek-v4-flash

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

Pith's one-line read Recurrent ground cues cut cross-view localization error to 1.57 m

desk verdict Strong benchmark paper with a genuinely new temporal attention design, but the field experiment needs a low-GPS baseline before the deployment claim lands. read the letter →

arxiv 2608.10660 v1 pith:B5XXU5HC submitted 2026-08-11 cs.CV cs.AI

Jiaping Wang , Shaobo Li , Zhen Wang This is my paper · ORCID
classification cs.CVcs.AI
keywords cross-viewlocalizationvisualtemporalcontextrecurrentattentionsatellite-groundmatchingautonomousdrivingsequencetwo-stage
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

This paper argues that cross-view visual localization for autonomous driving should not match each ground frame to the satellite map in isolation; instead, historical context from earlier frames should be injected into the coarse ground features before matching. It proposes a recurrent cross-frame module that lets each current frame retrieve cues from the previous recurrent state while preserving its own spatial layout, followed by a two-stage coarse-to-fine localization. On the CVIS benchmark the full model lowers mean localization error from 3.80 m to 1.57 m and raises R@1 m from 8.14% to 40.22%, with strong transfer to KITTI-CVL and zero-shot real-vehicle results. The paper's central claim is that temporal context injected at the right place, into coarse ground features before candidate-region classification, is what drives the large accuracy gain.

What carries the argument

The central mechanism is the recurrent cross-frame context enhancement module. It treats the current frame's coarse ground feature as the Query (with its own spatial positional encoding), the previous recurrent state as the Key and Value (with a separate positional encoding), and updates the state through cross-frame attention plus a residual feed-forward update so the current observation remains the main representation. This lets stable road structures and scene context propagate frame by frame into the coarse-feature stream, sharpening the satellite candidate-region distribution before fine offset regression. The two-stage localization then uses the enhanced coarse features for top-K candidate masking and intermediate-layer fine features for offset estimation.

What would settle it

Run the same model on CVIS sequences re-sampled with wider spacing (e.g., 16 m or 24 m between frames): if the temporal module stops improving over the no-temporal variant, or if mean error fails to rise, the continuity assumption underpinning the recurrent state is refuted.

Watch

Extended reading notes

Core claim

The central claim is that recurrent temporal context modeling on coarse ground features substantially improves cross-view sequential localization accuracy. The framework extracts hierarchical DINOv2 features from one satellite map and six consecutive ground images, enhances only the coarse ground feature stream with a recurrent cross-frame attention module, then performs satellite candidate-region classification on the enhanced features and candidate-conditioned offset regression on fine features. The authors report that on CVIS this design reduces mean error from 3.80 m (TACV) to 1.57 m and R@1 m from 8.14% to 40.22%; the same model transfers directly to KITTI-CVL with 2.61 m mean error and improves to 2.27 m with fine-tuning. A zero-shot field experiment on a real vehicle reports 2.84 m mean error and 96.86% R@5 m using low-accuracy GPS only to crop satellite maps. The ablation shows the temporal module contributes the largest single gain, from 4.96 m to 1.57 m mean error.

Load-bearing premise

The temporal enhancement assumes that consecutive ground frames, spaced about 8 m apart, share enough visual and geometric content that the previous recurrent state, computed from coarse DINOv2 features, provides useful retrieval context for the current frame.

Editorial extensions

If this is right

  • On CVIS, moving from independent frame matching to recurrent temporal enhancement cuts mean error from 3.80 m to 1.57 m and raises R@1 m from 8.14% to 40.22%.
  • The gain is not merely a stronger backbone: replacing TACV's extractor with DINOv2 only reaches 2.84 m and 15.95% R@1 m, while the full model reaches 1.57 m and 40.22%.
  • The model transfers across datasets: direct transfer to KITTI-CVL achieves 2.61 m mean error, and fine-tuning lowers it to 2.27 m with 35.69% R@1 m.
  • Zero-shot deployment on a real vehicle with low-accuracy GPS crops achieves 2.84 m mean error and 96.86% R@5 m across nine urban scenarios.
  • Temporal context improves Stage-1 candidate quality: under top-64 masking (17.73% of the search space) ground-truth grid coverage rises from 99.95% to 99.98%, and top-5 coverage from 83.33% to 98.92%.

Reading between the lines

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

  • The asymmetric Query-current / Key-Value-history design suggests a general recipe for sequence-to-map retrieval: use the current observation to decide what history to consult, rather than pooling or averaging frames; this could transfer to other cross-view tasks such as visual place recognition or drone-to-satellite localization.
  • Because the improvement concentrates in the coarse candidate stage, the method should benefit any downstream fine refiner that is conditioned on a candidate mask; replacing the offset regressor with a different localizer may preserve most of the gain.
  • A testable extension is to vary frame spacing: if the recurrent state is computed from frames spaced farther than 8 m, the overlap assumption should break and accuracy should degrade; the current experiments do not probe this boundary.
  • The real-vehicle experiment uses low-accuracy GPS only to crop the satellite map, suggesting the method could be combined with a coarse GNSS prior in a larger pipeline; the paper does not quantify how sensitive the result is to the crop-center error.
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

3 major / 4 minor

Summary. This paper proposes a recurrent spatio-temporal context module for sequential cross-view visual localization. A frozen pretrained DINOv2 extracts coarse and fine features from satellite and ground images; a recurrent cross-frame attention block updates the coarse ground feature of each frame using the previous recurrent state; a two-stage head then classifies candidate satellite grids and regresses local offsets. On CVIS the method reports mean error 1.57 m and R@1m 40.22%, outperforming TACV and TACV+DINOv2; on KITTI-CVL it reports 2.61 m mean error with direct transfer and 2.27 m after fine-tuning; a zero-shot field experiment reports 2.84 m mean error. The central claim is that injecting temporal context into coarse ground features before candidate classification drives the large gain.

Significance. If the results hold, the proposed method is a notable step for cross-view sequential localization, especially the 40.22% R@1m on CVIS compared with 8.14% for TACV. The paper deserves credit for comparing against TACV with the same DINOv2 backbone, for ablating the temporal module and multi-level features separately, and for evaluating zero-shot on a real vehicle. The main missing evidence is a control for the GPS prior in the field experiment and full reproducibility details; neither affects the internal consistency of the benchmark comparisons, but both are needed before the deployment claim can be accepted.

major comments (3)
  1. [Section IV.B, Table 9] The field experiment reports visual-model errors against RTK but never reports the errors of the low-accuracy GPS fixes that define the satellite-map crop centers. Because the low-GPS module is described as meter-level, a control that simply outputs the per-frame low-GPS fix could plausibly match or beat the reported 2.84 m mean error and 96.86% R@5 m. Please add a low-GPS-only baseline with the same metrics and discuss where the visual model improves on it; otherwise the abstract/conclusion claim of robust real-world deployment is not supported by Table 9. Additionally, Fig. 15 shows RTK can be locally incorrect, so quantitative errors in Table 9 may be corrupted by ground-truth drift; please either restrict numerical evaluation to trusted RTK segments or present these cases qualitatively.
  2. [Section III.C, Table 3, Eqs. (7), (11)] The reproducibility of the central empirical claims is incomplete. The loss weights lambda_cls and lambda_reg in Eq. (11), the optimizer and its schedule, the number of training epochs, and the inference top-K used in the reported results are not given; Table 7 reports coverage for top-1 through top-64 but does not state which K produced Tables 4-6 and 9. Since no code or trained model is provided, these values are needed to reproduce the experiments. Please report them, and if possible include standard deviations or confidence intervals from multiple seeds for Tables 4, 5, 6, and 9.
  3. [Section III.D, Tables 4 and 5] The no-temporal ablation is coherent, but its error distribution should be reported. On the CVIS test set, Ours w/o Temporal has mean error 5.92 m yet R@1m of 24.44%, while TACV+DINOv2 has mean error 2.84 m and R@1m of 15.95%; on the KITTI-CVL test set, Ours w/o Temporal has mean error 14.38 m but R@1m of 9.59%, while TACV has mean error 3.57 m and R@1m of 0.00%. These combinations indicate that the means are dominated by long error tails and that the relative ordering of methods can differ between mean and recall. Please report error percentiles or cumulative error distributions so the magnitude of the temporal-context gain is not an artifact of a few catastrophic frames.
minor comments (4)
  1. [Tables 4, 5, and 9] Please clarify whether the metrics are computed per frame or per sequence; the field table reports 'Sequences' counts, and the text alternates between frame-level and trajectory-level language.
  2. [Section III.C, Table 3] Table 3 lists 'Satellite grid 19x19' but not the grid cell size; report it for CVIS and KITTI-CVL so the two-stage localization can be interpreted.
  3. [Section II.F, Eqs. (12) and (13)] The notation is inconsistent: Eq. (13) writes t-hat with an n_gt superscript while the surrounding text uses (b,t) indices, and Eq. (12) uses double superscripts for y. Please unify the notation.
  4. [Figures 8, 10, and 15] Figure captions use 'GT trajectory', 'GT', and 'Green dots denote GT' interchangeably; please unify and state explicitly in Fig. 15 that the green dots are RTK positions.

Circularity Check

0 steps flagged · score 0.0 of 10

No circularity: benchmark comparisons and temporal-context ablations are self-contained empirical evaluations.

full rationale

The paper's derivation chain is supervised empirical learning. The temporal enhancement module (Eqs. 1-4) is a defined network architecture, not derived from the target results; the losses (Eqs. 11-13) supervise grid classification and offset regression against held-out ground truth. The central claims are supported by test-set comparisons on CVIS (Table 4) and KITTI-CVL (Table 5) against external baselines, plus ablations (Table 6) that isolate multi-level features and temporal attention. No parameter is fitted to the test set and then reported as a prediction; the zero-shot field experiment (Section IV) uses no training or fine-tuning on field data. There are no load-bearing self-citations: references are to external prior work, and DINOv2 is an externally pretrained backbone. The only evident evaluation gap is that the real-vehicle experiment (Section IV, Table 9) does not report the error of the low-accuracy GPS positions that define the satellite-map crop centers, so a trivial GPS-prior baseline is missing; this is a validation or control concern, not circularity, because the visual model's predictions are not algebraically or statistically forced by the low-GPS input. Hyperparameters chosen on the validation split are standard practice and do not make the reported test results definitionally circular.

Assumptions & free parameters 4 free parameters · 4 assumptions · 0 invented entities

No new physical entities are proposed. The method introduces a recurrent internal state and a top-K mask as architectural constructs, not independent entities with falsifiable handles. The main hidden cost is the set of hyperparameters above, none of which are released with fitted values.

free parameters (4)
  • Sequence length T = 6
    Six temporally ordered frames are sampled per sequence; all temporal context and latency numbers depend on this hand-chosen value (Section III.A, III.C).
  • Loss weights lambda_cls and lambda_reg = not reported
    The total objective in Eq. 11 requires balancing weights, but their values are not given, so the reported accuracy depends on an unspecified tuning choice.
  • Top-K candidate mask size = not reported
    Stage 2 offset regression is restricted to the top-K coarse grids (Eq. 7); the coverage analysis in Table 7 shows accuracy varies strongly with K, yet the K used for training and inference is not stated.
  • Satellite grid resolution = 19 x 19
    The coarse localization discretizes the satellite map into a 19 by 19 grid (Table 3); this discretization determines the offset regression range and therefore the achievable accuracy.
assumptions (4)
  • domain assumption Pretrained DINOv2 features transfer to cross-view matching across CVIS, KITTI, and real vehicle imagery
    The model relies on DINOv2 features learned on other data without fine-tuning the backbone except a small learning rate; if the feature distribution shifts, reported transfer results would not hold (Section II.C, III.C).
  • domain assumption Ground and satellite views of the same location are visually matchable from appearance alone
    The cross-view matching paradigm assumes sufficient shared structure between street-level and overhead imagery to supervise grid classification and offset regression (Section II.E).
  • ad hoc to paper The constructed KITTI-CVL sequences are a valid proxy for sequential cross-view localization
    Sequences are generated by sampling every 8 m and discarding those with fewer than six frames, then starting new sequences from midpoints; this self-created split defines the transfer benchmark (Section III.A).
  • domain assumption RTK GPS provides reliable ground truth for the real-vehicle experiments
    Errors are computed against RTK trajectories, while the paper also states RTK drifts in high-rise and obstructed scenes, making the ground truth partly self-contradictory (Section IV, Fig. 15).

how reviews work

0 comments
Cite this review

Pith. "Pith review of Cross-View Sequential Visual Localization with Spatio-Temporal Context Modeling for Autonomous Driving." pith.science (2026). https://pith.science/paper/B5XXU5HC

@misc{pith2026260810660,
  author       = {Pith},
  title        = {Pith review of: Cross-View Sequential Visual Localization with Spatio-Temporal Context Modeling for Autonomous Driving},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/B5XXU5HC}},
  note         = {Machine review of arXiv:2608.10660}
}
read the original abstract

Continuous and reliable localization is essential for autonomous driving. Cross-view visual localization matches ground images with satellite maps, providing complementary localization cues for pipelines that depend on Global Navigation Satellite System (GNSS) signals and high-definition (HD) maps. Most existing cross-view visual localization methods process each frame independently, leaving temporal information underused and limiting accuracy under dynamic occlusion, illumination variation, and repetitive textures. This study proposes a temporal-context-enhanced framework for cross-view sequence visual localization. The proposed recurrent cross-frame module aggregates historical context from the previous state to enhance the coarse ground feature of each current frame. These enhanced features facilitate satellite candidate-region classification, while hierarchical fine-grained features enable precise local offset estimation. On the CVIS dataset, the proposed method reduces mean localization error from 3.80 m to 1.57 m and increases R@1 m from 8.14% to 40.22%. Direct transfer to KITTI-CVL achieves a mean error of 2.61 m, with target-domain fine-tuning further reducing the mean error to 2.27 m. Zero-shot field experiments on a real-world vehicle achieve a mean error of 2.84 m and R@5 m of 96.86%. These results demonstrate that temporal context enhancement significantly improves cross-view localization accuracy and supports robust deployment on public benchmarks and real-world roads.

Figures

Figures reproduced from arXiv: 2608.10660 by the authors.

Figure 1
Figure 1. Illustration of cross‑view localization. Most existing cross-view localization methods still treat road video as independent single-frame matching tasks [15]. Recent studies have improved single-frame localization through shared representation learning, cross-view fusion, local region alignment, and fine-grained position regression [16][17]. Shi et al. [18] reformulate retrieval-based localization as neural camera p… view at source ↗
Figure 2
Figure 2. Overview of the proposed sequential localization framework. sequence 𝑆𝑔 = {𝐼𝑔,𝑡 } 𝑡=0 𝑇−1 , where 𝐼𝑔,𝑡 ∈ ℝ 𝐻𝑔×𝑊𝑔×3 denotes the image captured by vehicle-mounted cameras at timestamp 𝑡, and a satellite map 𝐼𝑠 ∈ ℝ 𝐻𝑠×𝑊𝑠×3 covering the driving corridor without being centered on any queried vehicle position, the task estimates the planar vehicle positions {(𝑥𝑡 , 𝑦𝑡 )}𝑡=0 𝑇−1 in the satellite coordinate system. To solve … view at source ↗
Figure 3
Figure 3. Architecture of the DINOv2-based hierarchical feature extraction. The satellite and ground hierarchical features are projected to a unified channel dimension 𝐷 through 1 × 1 convolution and then flattened along spatial dimensions. The projected ground-sequence features are further refined by the temporal enhancement module, and the enhanced sequence features are aggregated and matched in the two-stage localization m… view at source ↗
Figures from the paper (12 more)
Figure 4
Figure 4. Figure 4: Illustration of the spatio-temporal enhancement structure [PITH_FULL_IMAGE:figures/full_fig_p004_4.png]
Figure 5
Figure 5. Figure 5: Architecture of the dual-stream coarse localization network. Stage 2. As shown in [PITH_FULL_IMAGE:figures/full_fig_p005_5.png]
Figure 6
Figure 6. Figure 6: Illustration of the mask-guided fine-grained localizer. Specifically, according to the coarse classification scores, the top-K grids are selected to construct a binary candidate mask: ℳ𝑡,𝑛 = { 1, 𝑛 ∈ 𝑇𝑜𝑝𝐾(𝑠𝑡 ), 0, 𝑜𝑡ℎ𝑒𝑟𝑤𝑖𝑠𝑒. (7) The candidate mask does not change the f…
Figure 7
Figure 7. Figure 7: Recall rate curves of different methods based on the CVIS dataset [PITH_FULL_IMAGE:figures/full_fig_p008_7.png]
Figure 8
Figure 8. Figure 8: visualizes localization results on two consecutive temporal clips. The predicted positions closely follow the ground-truth positions in spatial distribution. Together, [PITH_FULL_IMAGE:figures/full_fig_p008_8.png]
Figure 9
Figure 9. Figure 9: Recall rate curves of different methods on the KITTI￾CVL dataset [PITH_FULL_IMAGE:figures/full_fig_p009_9.png]
Figure 10
Figure 10. Figure 10: visualizes localization results on continuous KITTI￾CVL temporal clips. The predicted positions are spatially close to the ground-truth positions in the visualized clips, indicating effective localization behavior in unseen road scenes. Together, [PITH_FULL_IMAGE:fig…
Figure 12
Figure 12. Figure 12: shows the vehicle used in the experiments. The vehicle is equipped with a front-facing camera, a high￾accuracy GPS device, and a low-accuracy GPS device to synchronously collect road images and vehicle positioning information. As reported in [PITH_FULL_IMAGE:figures/…
Figure 11
Figure 11. Figure 11: further visualizes the Stage 1 probability distributions. Without temporal context, the response distribution tends to spread across multiple visually similar road regions. After the temporal context is introduced, high￾confidence responses become more concentrated ar…
Figure 13
Figure 13. Figure 13: Driving route in the field experiments. For each ground-image sequence, this study uses the low￾accuracy GPS position of the center frame as the crop center for the satellite image and retrieves the corresponding satellite image from Google Maps. Satellite images are …
Figure 14
Figure 14. Figure 14: visualizes localization results across different real￾vehicle scenarios. For internal roads, urban arterial roads, and side roads under viaducts, our model's results are generally close to the GT, demonstrating the strong usability in the real world. In roundabouts, e…
Figure 15
Figure 15. Figure 15: Qualitative cases with local inconsistencies between RTK-based GT and visible road geometry. Green dots denote results from RTK, and blue diamonds denote our model's results. Overall, [PITH_FULL_IMAGE:figures/full_fig_p012_15.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

32 extracted references · 27 canonical work pages

  1. [1]

    Real -time performance -focused localization techniques for autonomous vehicle: A review,

    Y. Lu, H. Ma, E. Smart, and H. Yu, “Real -time performance -focused localization techniques for autonomous vehicle: A review,” IEEE Trans. Intell. Transp. Syst., vol. 23, no. 7, pp. 6082 –6100, Jul. 2022, doi: 10.1109/TITS.2021.3077800

    arXiv 2022

  2. [2]

    EdgeUNet: Edge -guided multi-loss network for drivable area and lane segmentation in autonomous vehicles,

    X. Sheng et al., “EdgeUNet: Edge -guided multi-loss network for drivable area and lane segmentation in autonomous vehicles,” IEEE Trans. Intell. Transp. Syst., vol. 26, no. 2, pp. 2530–2542, Nov. 2024

    work page 2024

  3. [3]

    LSMCL: Long -term static mapping and cloning localization for autonomous robot navigation using 3D LiDAR in dynamic environments,

    Y. C. Lee, “LSMCL: Long -term static mapping and cloning localization for autonomous robot navigation using 3D LiDAR in dynamic environments,” Expert Syst. Appl., vol. 241, 2024, Art. no. 122688

    work page 2024

  4. [4]

    Robust global localization for urban autonomous vehicles via 3D geometric -enhanced visual place recognition,

    J. Shang, Y. Liu, Y. Xu, J. Xiao, and D. Ma, "Robust global localization for urban autonomous vehicles via 3D geometric -enhanced visual place recognition," IEEE Trans. Intell. Transp. Syst., vol. 26, no. 11, pp. 18553- 18567, 2025

    work page 2025

  5. [5]

    End -to-end autonomous driving: Challenges and frontiers,

    L. Chen, P. Wu, K. Chitta, B. Jaeger, A. Geiger, and H. Li, “End -to-end autonomous driving: Challenges and frontiers,” IEEE Trans. Pattern Anal. Mach. Intell., vol. 46, no. 12, pp. 10164 –10183, Dec. 2024, doi: 10.1109/TPAMI.2024.3435937

    arXiv 2024

  6. [6]

    Cross-view geo-localization: A survey,

    A. Durgam, S. Paheding, V. Dhiman, and V. Devabhaktuni, “Cross-view geo-localization: A survey,” IEEE Access, vol. 12, pp. 192028 –192050, 2024

    work page 2024

  7. [7]

    View consistent purification for accurate cross -view localization,

    S. Wang, Y. Zhang, A. Perincherry, A. Vora, and H. Li, “View consistent purification for accurate cross -view localization,” in Proc. IEEE/CVF Int. Conf. Comput. Vis. (ICCV), Paris, France, 2023, pp. 8197–8206

    work page 2023

  8. [8]

    Cross-view matching for vehicle localization by learning geographically local representations,

    Z. Xia, O. Booij, M. Manfredi, and J. F. P. Kooij, “Cross-view matching for vehicle localization by learning geographically local representations,” IEEE Robot. Autom. Lett., vol. 6, no. 3, pp. 5921–5928, Jul. 2021

    work page 2021

Show all 32 references
  1. [9]

    Self -supervised cross-view graph search framework for ground -to-satellite geo - localization,

    K. Xie, W. Zhou, X. Huang, H. Guan, and F. Yulong, “Self -supervised cross-view graph search framework for ground -to-satellite geo - localization,” IEEE Trans. Geosci. Remote Sens., early access, Aug. 25, 2025

    2025

  2. [10]

    SliceMatch: Geometry-guided aggregation for cross -view pose estimation,

    T. Lentsch, Z. Xia, H. Caesar, and J. F. P. Kooij, “SliceMatch: Geometry-guided aggregation for cross -view pose estimation,” in Proc. IEEE/CVF Conf. Comput. Vis. Pattern Recognit. (CVPR), Vancouver, BC, Canada, 2023, pp. 17225–17234

    2023

  3. [11]

    BevSplat: Resolving height ambiguity via feature-based Gaussian primitives for weakly -supervised cross -view localization,

    Q. Wang, S. Wu, and Y. Shi, “BevSplat: Resolving height ambiguity via feature-based Gaussian primitives for weakly -supervised cross -view localization,” 2025, arXiv:2502.09080

    2025

  4. [12]

    AnyLoc: Towards universal visual place recognition,

    N. Keetha et al., “AnyLoc: Towards universal visual place recognition,” IEEE Robot. Autom. Lett., vol. 9, no. 2, pp. 1286–1293, Dec. 2023

    2023

  5. [13]

    Benchmarking the robustness of cross-view geo- localization models,

    Q. Zhang and Y. Zhu, “Benchmarking the robustness of cross-view geo- localization models,” in Proc. Eur. Conf. Comput. Vis. (ECCV), Milan, Italy, 2024, pp. 36–53

    2024

  6. [14]

    MS2 -CL: Multi -scale self - supervised learning for camera to LiDAR cross-modal place recognition,

    W. Liu, L. Ma, X. Zhuang, and Z. Deng, “MS2 -CL: Multi -scale self - supervised learning for camera to LiDAR cross-modal place recognition,” Sensors, vol. 26, no. 5, 2026, Art. no. 1561

    2026

  7. [15]

    Video2BEV: Transforming drone videos to BEVs for video -based geo -localization,

    H. Ju, S. Huang, S. Liu, and Z. Zheng, “Video2BEV: Transforming drone videos to BEVs for video -based geo -localization,” in Proc. IEEE/CVF Int. Conf. Comput. Vis. (ICCV), 2025, pp. 27073–27083

    2025

  8. [16]

    Cross -view object geo-localization in a local region with satellite imagery,

    Y. Sun et al., “Cross -view object geo-localization in a local region with satellite imagery,” IEEE Trans. Geosci. Remote Sens., vol. 61, pp. 1 –6, 2023

    2023

  9. [17]

    Boosting 3 -DoF ground-to-satellite camera localization accuracy via geometry -guided cross-view transformer,

    Y. Shi, F. Wu, A. Perincherry, A. Vora, and H. Li, “Boosting 3 -DoF ground-to-satellite camera localization accuracy via geometry -guided cross-view transformer,” in Proc. IEEE/CVF Int. Conf. Comput. Vis. (ICCV), Paris, France, 2023, pp. 21516–21526

    2023

  10. [18]

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

    Y. Shi and H. Li, “Beyond cross -view image retrieval: Highly accurate vehicle localization using satellite image,” in Proc. IEEE/CVF Conf. Comput. Vis. Pattern Recognit. (CVPR), New Orleans, LA, USA, 2022, pp. 17010–17020

    2022

  11. [19]

    Visual cross -view metric localization with dense uncertainty estimates,

    Z. Xia, O. Booij, M. Manfredi, and J. F. P. Kooij, “Visual cross -view metric localization with dense uncertainty estimates,” in Proc. Eur. Conf. Comput. Vis. (ECCV), Tel Aviv, Israel, 2022, pp. 90–106

    2022

  12. [20]

    Sample4Geo: Hard negative sampling for cross-view geo-localisation,

    F. Deuser, K. Habel, and N. Oswald, “Sample4Geo: Hard negative sampling for cross-view geo-localisation,” in Proc. IEEE/CVF Int. Conf. Comput. Vis. (ICCV), Paris, France, 2023, pp. 16847–16856

    2023

  13. [21]

    Fine -grained cross- view geo-localization using a correlation-aware homography estimator,

    X. Wang, R. Xu, Z. Cui, Z. Wan, and Y. Zhang, “Fine -grained cross- view geo-localization using a correlation-aware homography estimator,” Adv. Neural Inf. Process. Syst., vol. 36, pp. 5301–5319, 2023

    2023

  14. [22]

    BEV -CV: Birds -eye-view transform for cross-view geo-localisation,

    T. Shore, S. Hadfield, and O. Mendez, “BEV -CV: Birds -eye-view transform for cross-view geo-localisation,” in Proc. IEEE/RSJ Int. Conf. Intell. Robots Syst. (IROS), Abu Dhabi, United Arab Emirates, 2024, pp. 11048–11055

    2024

  15. [23]

    Cross -attention between satellite and ground views for enhanced fine -grained robot geo -localization,

    D. Yuan, F. Maire, and F. Dayoub, “Cross -attention between satellite and ground views for enhanced fine -grained robot geo -localization,” in Proc. IEEE/CVF Winter Conf. Appl. Comput. Vis. (WACV), Waikoloa, HI, USA, 2024, pp. 1249–1256

    2024

  16. [24]

    FG^2: Fine -grained cross -view localization by fine-grained feature matching,

    Z. Xia and A. Alahi, “FG^2: Fine -grained cross -view localization by fine-grained feature matching,” in Proc. IEEE/CVF Conf. Comput. Vis. Pattern Recognit. (CVPR), 2025, pp. 6362–6372

    2025

  17. [25]

    RSO -SLAM: A robust semantic visual SLAM with optical flow in complex dynamic environments,

    L. Qin, C. Wu, Z. Chen, X. Kong, Z. Lv, and Z. Zhao, “RSO -SLAM: A robust semantic visual SLAM with optical flow in complex dynamic environments,” IEEE Trans. Intell. Transp. Syst., vol. 25, no. 10, pp. 14669–14684, Oct. 2024, doi: 10.1109/TITS.2024.3402241

    2024

  18. [26]

    A deep analysis of visual SLAM methods for highly automated and autonomous vehicles in complex urban environment,

    K. Wang, G. Zhao, and J. Lu, “A deep analysis of visual SLAM methods for highly automated and autonomous vehicles in complex urban environment,” IEEE Trans. Intell. Transp. Syst., vol. 25, no. 9, pp. 10524–10541, Sep. 2024, doi: 10.1109/TITS.2024.3379993

    2024

  19. [27]

    CVLNet: Cross -view semantic correspondence learning for video -based camera localization,

    Y. Shi, X. Yu, S. Wang, and H. Li, “CVLNet: Cross -view semantic correspondence learning for video -based camera localization,” in Proc. Asian Conf. Comput. Vis. (ACCV), Macao, China, 2022, pp. 123–141

    2022

  20. [28]

    GAMa: Cross -view video geo - localization,

    S. Vyas, C. Chen, and M. Shah, “GAMa: Cross -view video geo - localization,” in Proc. Eur. Conf. Comput. Vis. (ECCV), Tel Aviv, Israel, 2022, pp. 440–456

    2022

  21. [29]

    Cross -view image sequence geo - localization,

    X. Zhang, W. Sultani, and S. Wshah, “Cross -view image sequence geo - localization,” in Proc. IEEE/CVF Winter Conf. Appl. Comput. Vis. (WACV), Waikoloa, HI, USA, 2023, pp. 2914–2923

    2023

  22. [30]

    Leveraging cross -view geo -localization with ensemble learning and temporal awareness,

    A. Ghanem et al., “Leveraging cross -view geo -localization with ensemble learning and temporal awareness,” PLOS ONE, vol. 18, no. 3, Mar. 2023, Art. no. e0283672, doi: 10.1371/journal.pone.0283672

    2023 doi

  23. [31]

    Temporal attention for cross -view sequential image localization,

    D. Yuan, F. Maire, and F. Dayoub, “Temporal attention for cross -view sequential image localization,” in Proc. IEEE/RSJ Int. Conf. Intell. Robots Syst. (IROS), Abu Dhabi, United Arab Emirates, 2024, pp. 7429–7436

    2024

  24. [32]

    Spatiotemporal contrastive learning for cross-view video localization in unstructured off-road terrains,

    Z. Deng et al., “Spatiotemporal contrastive learning for cross-view video localization in unstructured off-road terrains,” 2025, arXiv:2506.05250

    2025 arXiv

Pith tools

Reviewed August 12, 2026 · model on record in the stance chip above.