Pith. sign in

REVIEW 3 major objections 6 minor 56 references

Unsupervised Multi-view UAV Image Geo-localization via Iterative Rendering

T0 review · 3 major / 6 minor · reviewed 2026-08-12 · deepseek-v4-flash

Pith's one-line read The paper claims that reconstructing a UAV scene as 3D Gaussians, rendering an orthographic satellite-like view, and iteratively refining the virtual camera pose against retrieved satellite candidates lets a frozen feature extractor…

desk verdict A genuinely training-free 3DGS rendering approach to UAV-to-satellite retrieval that beats a 2020 supervised baseline; the iterative refinement is novel but its reach beyond initial top-K retrieval is unmeasured. read the letter →

arxiv 2411.14816 v1 pith:UHNHKT6A submitted 2024-11-22 cs.CV cs.ROeess.IV

classification cs.CVcs.ROeess.IV
keywords cross-viewgeo-localizationUAVimageretrieval3DGaussianSplattingorthographicrenderingiterativecameraposerefinementview-consistentfeaturefusionunsupervisedsatellitematching
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 tries to establish that cross-view UAV geo-localization can be solved without any paired drone-satellite training by lifting the query scene into 3D and generating the satellite view from the drone images themselves. It reconstructs the scene with 3D Gaussian Splatting, renders an orthographic top-down image, and uses a pre-trained foundation model to retrieve a matching satellite tile. A loop then refines the virtual camera pose using retrieved candidates and fuses the rendered views through self- and cross-view consistency. On the University-1652 benchmark the method reports 76.57% top-1 recall, above the 69.33% of the supervised baseline, and on SUES-200 it reports 73.0-76.5% top-1 recall without any task-specific training. If correct, this would mean UAV geo-localization can transfer to new regions without retraining or region-specific data.

What carries the argument

The central object is the orthographic virtual camera, whose rotation is aligned perpendicular to a RANSAC-fitted ground plane and whose translation and scale are chosen by maximizing the rendering area while penalizing blank regions. The supporting machinery is 3D Gaussian Splatting with an orthogonal-projection Jacobian, which renders the reconstructed scene as a satellite-like image, together with an SE(3) interpolation step that moves the virtual camera toward geometrically verified satellite candidates and a consistency-guided fusion that weights each re-rendered feature by its similarity to the previous rendered view and to its corresponding gallery image. This combination converts the cross-view retrieval problem into a self-aligning rendering loop that needs no learned viewpoint-invariant features.

What would settle it

Take a set of scenes and run the pipeline with the initial pose deliberately corrupted by increasing rotation or scale errors. If the final top-1 recall drops sharply once the true satellite patch falls outside the top-K candidates at the first retrieval step, and if later iterations cannot recover those cases, the initialization assumption is confirmed as load-bearing. A direct version is to report the recall of the T=0 retrieval alone: any query whose correct tile is absent from its top-K cannot be saved by iteration, so the final R@1 is bounded by that recall.

Watch

Extended reading notes

Core claim

The central claim is that an unsupervised, training-free drone-to-satellite retrieval pipeline can match or beat supervised methods by replacing the view-discrepancy problem with a rendering problem. The method first runs structure-from-motion on the multiple oblique drone images, initializes 3D Gaussian primitives from the sparse points, and fits a ground plane by RANSAC to define an initial orthographic virtual camera. After training the Gaussians, it renders a bird's-eye image, extracts a global descriptor with a frozen foundation model and GeM aggregation, and then repeatedly retrieves the top-K satellite candidates, verifies them by feature matching and pose consistency, interpolates the camera pose in SE(3) toward the verified candidates, re-renders, and fuses the new features with weights given by self-view consistency (agreement with the previous rendered view) and cross-view consistency (agreement with the candidate satellite image). The paper reports that two iterations of this refinement raise University-1652 top-1 recall from 57.43% to 76.57%, exceeding the supervised University-1652 baseline of 69.33%, and that on SUES-200 the method reaches 76.50% top-1 at 300-meter height without fine-tuning.

Load-bearing premise

The method only works if the initial virtual camera pose and scale, estimated from the ground plane and the blank-area objective, put the true satellite tile inside the top-K retrieval list; otherwise the iterative refinement has no correct candidate to lock onto.

Editorial extensions

If this is right

  • Annotated and paired drone-satellite training data becomes unnecessary for UAV geo-localization in a new region; only multi-view drone images and a geo-tagged satellite database are required.
  • Retrieval accuracy improves as more drone views are used, with a sharp drop below roughly 20 views and saturation by 50 views, indicating a view-count threshold for reliable scene reconstruction and rendering.
  • The iterative refinement converges in about two iterations, so the added computational cost is bounded and the pipeline remains practical for offline or near-real-time geo-localization.
  • A poor initial camera pose or scale can cause the refinement to lock onto a wrong satellite candidate, so the initialization stage is as load-bearing as the learned components in supervised methods.
  • The rendered orthographic images can serve as updated satellite gallery tiles, which could refresh stale or outdated satellite imagery in a region without new satellite capture.

Reading between the lines

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

  • The pipeline's core mechanism is not UAV-specific: any multi-view capture of a static scene could be lifted to 3D and rendered from an overhead viewpoint, so the same rendering-and-refine scheme could plausibly extend to ground-level or handheld multi-view imagery.
  • The consistency scores used for feature fusion could double as a confidence signal for active view selection, prompting a drone to fly to a better vantage point when the top candidate agreement is low.
  • The paper's observation that overlapping satellite patches cause metric ambiguity suggests that retrieval benchmarks should adopt a distance-weighted or cover-aware evaluation, since two neighboring tiles can both be correct for one drone scene.
  • A stress test with synthetic scenes of known geometry and deliberately corrupted initial poses would isolate how much of the reported accuracy comes from the initialization quality versus the refinement loop.
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 an unsupervised UAV-to-satellite cross-view geo-localization pipeline. Given multiple oblique drone images of a scene, the method reconstructs a 3D Gaussian Splatting model, estimates an initial virtual orthographic camera from a RANSAC ground-plane fit, renders a satellite-like image, retrieves satellite candidates using frozen DINOv2 global features, and then iteratively refines the virtual camera pose and fuses rendered candidate features using self- and cross-view consistency. The method does not fine-tune or train on paired data. Experiments on University-1652 and SUES-200 report R@1 of 76.57 and 76.50, respectively, outperforming the zero-shot AnyLoc baseline and reaching results comparable to the included supervised baselines.

Significance. The paper combines 3DGS-based rendering with a frozen foundation model to avoid training, which is a valuable direction for cross-region generalization in CVGL. The ablations in Tables III-V and Figure 11 show clear, consistent gains from rendering, iterative refinement, and consistency-guided fusion, and the failure cases in Figure 6 are discussed honestly. If the iterative refinement's improvement is robust beyond re-ranking already-correct candidates, the method is a meaningful contribution. However, the central empirical claim is not fully substantiated because the paper does not quantify how often the true satellite is in the initial top-K candidate set, and the supervised comparison is limited to older baselines.

major comments (3)
  1. [III-B, Eqs. (9)-(10), Table IV] The iterative refinement is the core contribution, but the paper never reports whether the true satellite is in the initial Top-K candidate set. Since candidate poses are derived from retrieved images and the fusion weights in Eq. (10) reward agreement with those candidates, a query whose true target is outside the initial Top-10 cannot be recovered by the loop. The observed improvement from R@1 57.43 at T=0 to 76.57 at T=2 could be entirely re-ranking within already-correct candidates. The authors should report T=0 R@10 and the conditional success rate split by whether the true target is in the initial candidate set, and ideally analyze how often the loop converges to a wrong candidate. Without this, the claim that iterative refinement 'progressively aligns' the query to the true satellite is not established.
  2. [Table I and Abstract] The abstract claims 'competitive performance with recent supervised methods', but the only supervised baselines on University-1652 and SUES-200 are Zheng et al. [15] from 2020 and Zhu et al. [5], respectively. No comparison with recent supervised CVGL methods (2021-2024) is provided, so the claim as stated is not supported. In addition, all results are single-run point estimates with no error bars or repeated runs; given the pipeline's randomness (random drone image selection in Eq. 7, RANSAC, 3DGS training), the authors should either report repeated-run statistics or justify why variance is negligible.
  3. [IV-F, Eq. (6), Fig. 6(a)] The virtual camera initialization is a known failure point (Fig. 6(a) and the Limitations paragraph), but the paper does not analyze the sensitivity of the method to initialization hyperparameters such as λ_m, s_w, s_h, or the RANSAC ground-plane estimate. Since the refinement loop depends on the initial rendered image being a reasonable retrieval query, an ablation or perturbed-initialization stress test is needed to support the method's claimed robustness across regions. At minimum, the authors should quantify how often Eq. (6) produces a 'failed initialization' like the one shown in Fig. 6(a).
minor comments (6)
  1. [III-A-2] Typo: 'Struction-from-motion' should be 'Structure-from-motion'.
  2. [III-B-1-b] The text defines the angular offset twice as Δ𝑑; the second definition should refer to Δ𝜃, as in the caption of Fig. 3.
  3. [III-B-2 and Fig. 4 caption] In the sentence before Eq. (10), 'to regular the feature updation' should be 'to regularize the feature update'.
  4. [Table II] The column header 'Fine-tuning Set' is ambiguous for the first row (Zhu et al., N/A); please clarify whether N/A means no fine-tuning or an empty fine-tuning set.
  5. [Fig. 11] The figure lacks a labeled horizontal axis; the caption should state explicitly that the x-axis is the fusion type (α, β, α+β).
  6. [IV-G 'Geographical Comparison'] The paragraph refers to 'GPS annotation' and a 'meter-level distance' metric, but the metric is not defined; specify how the distance between retrieved and true satellite patches is computed.

Circularity Check

1 steps flagged · score 3.0 of 10

Iterative refinement fuses candidate-derived renderings into the query, so the T=2 retrieval gain is partly self-referential; the core rendering pipeline is otherwise self-contained.

  1. fitted input called prediction [Section III-B, Eqs. 8-10 (Candidate Camera Pose Update and View Consistency-guided Feature Fusion); final retrieval in Section III-A.4]
    "The Top-K high similarity features of the retrieved satellite images are selected as candidates to perform feature matching and relative camera pose estimation. ... The consistency of cross-view calculates the similarity β between each e_{t,k} and the feature r_k of its corresponding image from the satellite gallery. ... e_t = λ_s e_{t-1}+(1-λ_s) Σ_k softmax_k(α_k β_k) e_{t,k}."

    By construction, the final query feature e_t is a convex combination of rendered-candidate features e_{t,k}, and each e_{t,k} is rendered from a virtual camera pose T_{t,k} that was estimated from the retrieved satellite candidate r_k. The fusion weight for candidate k is β_k = e_{t,k}^T r_k, i.e. the similarity between the candidate-derived rendering and that same candidate. The final retrieval score for candidate k is e_t^T r_k, which contains a term proportional to β_k times the softmax weight plus other positive contributions from the same candidate. Thus each top-K candidate's own similarity to the query directly inflates its final rank.

full rationale

The paper's core T=0 pipeline is self-contained: the 3D Gaussian Splatting model is trained only on the drone-view images of the query scene, the virtual camera is initialized from a RANSAC ground-plane fit and the blank-area objective in Eq. 6, and the feature extractor is a frozen DINOv2 model. No ground-truth labels or model fine-tuning are used, so the central claim of unsupervised geo-localization is not forced by fitting a parameter to the target data. The load-bearing caveat is the iterative refinement loop: satellite candidates retrieved at each step are used to estimate candidate camera poses (Eq. 8), render new query images, and then these candidate-derived renderings are fused into the query feature with weights β_k that measure similarity to the same candidates (Eqs. 9-10). This makes each candidate's final score partially self-referential, and the paper does not report the initial top-10 recall that would separate genuine refinement gains from re-ranking. However, the T=0 results are reported honestly, failure cases are shown in Fig. 6(a), and the paper makes no load-bearing self-citation or imported uniqueness claim. The circularity is partial, not total, so a score of 3 is appropriate rather than a higher score.

Assumptions & free parameters 9 free parameters · 6 assumptions · 0 invented entities

No new physical entities are introduced. The free parameters are hand-set hyperparameters or per-scene optimization quantities. The axioms are standard reconstruction and feature-transfer assumptions that the paper relies on without independent verification. The most load-bearing of these is the assumption that feature matching between a synthetic rendered orthographic image and a real satellite image produces valid relative poses.

free parameters (9)
  • a (SE(3) interpolation weight) = 0.8
    Hand-set in Eq. 8; controls how far the virtual camera moves toward each candidate pose per iteration.
  • lambda_m (blank-area penalty) = 100
    Hand-set in Eq. 6; balances rendered area against empty regions; affects initialization quality.
  • lambda_s (temporal fusion weight) = 0.5
    Hand-set in Eq. 10; balances previous scene feature and newly fused candidate features.
  • Top-K candidates = 10
    Hand-set; number of retrieved satellite patches used for candidate pose estimation and fusion.
  • N_m (matching inlier threshold) = 50
    Hand-set; minimum number of matched feature inliers for a candidate satellite image to survive matching verification.
  • T (number of refinement iterations) = 2
    Hand-set; Table IV shows R@1 saturates at T=2; the method stops after two iterations.
  • 3DGS training iterations = 7000
    Hand-set; used for every scene reconstruction; affects render quality.
  • render/reference resolution = 384x384
    Hand-set; used for all rendered and satellite images.
  • virtual camera scale factors s_w, s_h = per scene from Eq. 6
    Optimized per scene by maximizing coverage while penalizing blank areas; directly controls rendered scale and is listed as a failure source in Fig. 6(a).
assumptions (6)
  • domain assumption COLMAP/SfM recovered camera poses and sparse points accurately represent the UAV scene.
    Section III-A2 relies on SfM [48] for camera poses and initial points; if sparse reconstruction is poor, Gaussian initialization fails.
  • domain assumption RANSAC ground-plane detection identifies the dominant ground plane of the scene.
    Section III-A3 uses it to set the top-down virtual camera rotation R0.
  • domain assumption DINOv2 frozen features, aggregated with GeM, are suitable for matching rendered orthographic views to real satellite images without task-specific training.
    Section III-A4 uses DINOv2/GeM; the entire unsupervised claim depends on this transfer.
  • domain assumption Feature matching between a rendered orthographic image and a satellite image yields a valid relative camera pose estimate.
    Section III-B1 uses matched feature inliers to estimate candidate camera poses; no independent validation is provided.
  • domain assumption 3D Gaussian Splatting can be trained from N_v oblique views and rendered with orthographic projection to approximate a satellite view.
    Section III-A2; this is the core method assumption and is not independently verified beyond the reported retrieval results.
  • domain assumption The true satellite image is present in the searchable satellite database.
    All retrieval experiments assume the correct geo-tagged satellite patch is in the gallery; this is a dataset assumption of University-1652 and SUES-200.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Unsupervised Multi-view UAV Image Geo-localization via Iterative Rendering." pith.science (2026). https://pith.science/paper/UHNHKT6A

@misc{pith2026241114816,
  author       = {Pith},
  title        = {Pith review of: Unsupervised Multi-view UAV Image Geo-localization via Iterative Rendering},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/UHNHKT6A}},
  note         = {Machine review of arXiv:2411.14816}
}
read the original abstract

Unmanned Aerial Vehicle (UAV) Cross-View Geo-Localization (CVGL) presents significant challenges due to the view discrepancy between oblique UAV images and overhead satellite images. Existing methods heavily rely on the supervision of labeled datasets to extract viewpoint-invariant features for cross-view retrieval. However, these methods have expensive training costs and tend to overfit the region-specific cues, showing limited generalizability to new regions. To overcome this issue, we propose an unsupervised solution that lifts the scene representation to 3d space from UAV observations for satellite image generation, providing robust representation against view distortion. By generating orthogonal images that closely resemble satellite views, our method reduces view discrepancies in feature representation and mitigates shortcuts in region-specific image pairing. To further align the rendered image's perspective with the real one, we design an iterative camera pose updating mechanism that progressively modulates the rendered query image with potential satellite targets, eliminating spatial offsets relative to the reference images. Additionally, this iterative refinement strategy enhances cross-view feature invariance through view-consistent fusion across iterations. As such, our unsupervised paradigm naturally avoids the problem of region-specific overfitting, enabling generic CVGL for UAV images without feature fine-tuning or data-driven training. Experiments on the University-1652 and SUES-200 datasets demonstrate that our approach significantly improves geo-localization accuracy while maintaining robustness across diverse regions. Notably, without model fine-tuning or paired training, our method achieves competitive performance with recent supervised methods.

Figures

Figures reproduced from arXiv: 2411.14816 by the authors.

Figure 1
Figure 1. Illustration of the proposed scene rendering approach for cross￾view geo-localization. We propose a multi-view rendering image regression and retrieval approach that uses multiple UAV-captured views of a scene to predict its geo-location by retrieving matching satellite images from a database. First, we establish a 3D field representation of the query scene from multiple oblique images. Then, we iteratively render t… view at source ↗
Figure 2
Figure 2. Overview of the rendering-based UAV geo-localization. Given multiple oblique input views, our method first predicts an initial sparse reconstruction and learns a 3D representation using 3DGS. A virtual camera is estimated to render the scene and extract features for matching with real satellite images. The virtual camera pose is then iteratively updated based on feature matching, enabling high-fidelity novel view sy… view at source ↗
Figure 3
Figure 3. Candidate camera pose selection. The black denotes the previous camera’s coordinate system. Inlier camera poses are marked in green, while outliers are marked in red. In the previous camera’s coordinate, Δ𝑑 is the x-y distance to the previous camera, and Δ𝜃, is the angular deviation relative to the z-axis in the previous camera. plane relative to the previous camera’s coordinates. Second, we calculate the angular of… view at source ↗
Figures from the paper (6 more)
Figure 4
Figure 4. Figure 4: The update of the rendered images. After extracting the global features of the rendered candidates, we refine the scene’s features using a view consistency fusion module. The module first computes the self-view consistency 𝛼 between the rendered candidates and the prev…
Figure 5
Figure 5. Figure 5: Illustration of rendered images and the corresponding true satellite images. The first two rows show the samples of the input drone images. The third row shows the rendered images of the query scene. The rendered image can align with the true satellite image targets (g…
Figure 6
Figure 6. Figure 6: The visualization of the failed initialization of rendered images. The top row depicts that failed initialization of the camera pose leads to distorted rendering and subsequently unsuccessful pose refinement. The bottom row highlights a poor initial camera pose, but af…
Figure 8
Figure 8. Figure 8: Rendered images with different numbers of the input drone view. As the number of input views increases, the rendered images become more aligned with the satellite images, resulting in higher fidelity in the novel views. results indicate a noticeable decline in performa…
Figure 9
Figure 9. Figure 9: Meter-level accuracies in University-1652. reliability and performance. At this threshold, the quality of the rendering stabilizes and the output is consistent with the target imagery. Although increasing the number of views beyond this point further enhances the rende…
Figure 11
Figure 11. Figure 11: Top-1 accuracies of different fusion types. We perform an exhaustive ablation to report the improvement of different combinations of view consistency regularization on the fusion module. Combining both view consistencies yields the highest top-1 accuracy overall, with…

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

56 extracted references · 52 canonical work pages

  1. [15]

    University-1652: A multi-view multi- source benchmark for drone-based geo-localization,

    Z. Zheng, Y . Wei, and Y . Yang, “University-1652: A multi-view multi- source benchmark for drone-based geo-localization,” in Proceedings of the 28th ACM International Conference on Multimedia, 2020, pp. 1395–

  2. [5]

    Sues-200: A multi-height multi-scene cross-view image benchmark across drone and satellite,

    R. Zhu, L. Yin, M. Yang, F. Wu, Y . Yang, and W. Hu, “Sues-200: A multi-height multi-scene cross-view image benchmark across drone and satellite,” IEEE Transactions on Circuits and Systems for Video Technology, vol. 33, no. 9, pp. 4825–4839, 2023. 1, 2, 6, 7

  3. [1]

    Terrain aided planetary uav lo- calization based on geo-referencing,

    X. Wan, Y . Shao, S. Zhang, and S. Li, “Terrain aided planetary uav lo- calization based on geo-referencing,” IEEE Transactions on Geoscience and Remote Sensing , vol. 60, pp. 1–18, 2022. 1

  4. [2]

    Improving social media geolocation for disaster response by using text from images and chatgpt,

    H. B. Firmansyah, V . Lorini, M. O. Mulayim, J. Gomes, and J. L. Fernandez-Marquez, “Improving social media geolocation for disaster response by using text from images and chatgpt,” in Proceedings of the 2024 11th Multidisciplinary International Social Networks Conference , 2024, pp. 67–72. 1

  5. [3]

    A uav vision system for airborne surveillance,

    M. Kontitsis, K. Valavanis, and N. Tsourveloudis, “A uav vision system for airborne surveillance,” in International Conference on Robotics and Automation, vol. 1, 2004, pp. 77–83 V ol.1. 1

  6. [4]

    Hashing for geo-localization,

    P. Ren, Y . Tao, J. Han, and P. Li, “Hashing for geo-localization,” IEEE Transactions on Geoscience and Remote Sensing , vol. 61, pp. 1–13,

  7. [6]

    Mccg: A convnext- based multiple-classifier method for cross-view geo-localization,

    T. Shen, Y . Wei, L. Kang, S. Wan, and Y .-H. Yang, “Mccg: A convnext- based multiple-classifier method for cross-view geo-localization,” IEEE Transactions on Circuits and Systems for Video Technology , vol. 34, no. 3, pp. 1456–1468, 2024. 1

  8. [7]

    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 IEEE/CVF International Conference on Computer Vision , 2023, pp. 16 847–16 856. 1

Show all 56 references
  1. [8]

    Learning generalized zero- shot learners for open-domain image geolocalization,

    L. Haas, S. Alberti, and M. Skreta, “Learning generalized zero- shot learners for open-domain image geolocalization,” arXiv preprint arXiv:2302.00275, 2023. 1

  2. [9]

    Enhancing cross-view geo-localization with domain alignment and scene consistency,

    P. Xia, Y . Wan, Z. Zheng, Y . Zhang, and J. Deng, “Enhancing cross-view geo-localization with domain alignment and scene consistency,” IEEE Transactions on Circuits and Systems for Video Technology, 2024, early access. 1

  3. [10]

    Cross-view geo-localization with layer- to-layer transformer,

    H. Yang, X. Lu, and Y . Zhu, “Cross-view geo-localization with layer- to-layer transformer,” in Advances in Neural Information Processing Systems, vol. 34, 2021, pp. 29 009–29 020. 1

  4. [11]

    Multiple- environment self-adaptive network for aerial-view geo-localization,

    T. Wang, Z. Zheng, Y . Sun, C. Yan, Y . Yang, and T.-S. Chua, “Multiple- environment self-adaptive network for aerial-view geo-localization,” Pattern Recognition, vol. 152, p. 110363, 2024. 1 12

  5. [12]

    Unleashing unlabeled data: A paradigm for cross-view geo-localization,

    G. Li, M. Qian, and G.-S. Xia, “Unleashing unlabeled data: A paradigm for cross-view geo-localization,” in Proceedings of the IEEE/CVF Con- ference on Computer Vision and Pattern Recognition, 2024, pp. 16 719– 16 729. 1, 2, 3

  6. [13]

    Learning cross-view visual geo-localization without ground truth,

    H. Li, C. Xu, W. Yang, H. Yu, and G.-S. Xia, “Learning cross-view visual geo-localization without ground truth,” IEEE Transactions on Geoscience and Remote Sensing , vol. 62, pp. 1–17, 2024. 1, 2

  7. [14]

    Adapting fine-grained cross- view localization to areas without fine ground truth,

    Z. Xia, Y . Shi, H. Li, and J. F. Kooij, “Adapting fine-grained cross- view localization to areas without fine ground truth,” arXiv preprint arXiv:2406.00474, 2024. 1

  8. [16]

    Distinctive image features from scale-invariant keypoints,

    D. G. Lowe, “Distinctive image features from scale-invariant keypoints,” International journal of computer vision , vol. 60, pp. 91–110, 2004. 2

  9. [17]

    Uav pose estimation using cross-view geolocalization with satellite imagery,

    A. Shetty and G. X. Gao, “Uav pose estimation using cross-view geolocalization with satellite imagery,” in International Conference on Robotics and Automation . IEEE, 2019, pp. 1827–1833. 2

  10. [18]

    Soft exemplar highlighting for cross-view image-based geo-localization,

    Y . Guo, M. Choi, K. Li, F. Boussaid, and M. Bennamoun, “Soft exemplar highlighting for cross-view image-based geo-localization,” IEEE Transactions on Image Processing, vol. 31, pp. 2094–2105, 2022. 2

  11. [19]

    Cvm-net: Cross-view matching network for image-based ground-to-aerial geo-localization,

    S. Hu, M. Feng, R. M. Nguyen, and G. H. Lee, “Cvm-net: Cross-view matching network for image-based ground-to-aerial geo-localization,” in Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition, 2018, pp. 7258–7267. 2

  12. [20]

    Bridging the domain gap for ground-to- aerial image matching,

    K. Regmi and M. Shah, “Bridging the domain gap for ground-to- aerial image matching,” in Proceedings of the IEEE/CVF International Conference on Computer Vision , 2019, pp. 470–479. 2, 3

  13. [21]

    Vigor: Cross-view image geo-localization beyond one-to-one retrieval,

    S. Zhu, T. Yang, and C. Chen, “Vigor: Cross-view image geo-localization beyond one-to-one retrieval,” in Proceedings of the IEEE/CVF Confer- ence on Computer Vision and Pattern Recognition, 2021, pp. 3640–3649. 2

  14. [22]

    Uavs in multimedia: Capturing the world from a new perspective,

    Z. Zheng, Y . Shi, T. Wang, J. Liu, J. Fang, Y . Wei, and T.-s. Chua, “Uavs in multimedia: Capturing the world from a new perspective,” in Proceedings of the 31th ACM International Conference on Multimedia Workshop, vol. 4, 2023. 2

  15. [23]

    Each part matters: Local patterns facilitate cross-view geo-localization,

    T. Wang, Z. Zheng, C. Yan, J. Zhang, Y . Sun, B. Zheng, and Y . Yang, “Each part matters: Local patterns facilitate cross-view geo-localization,” IEEE Transactions on Circuits and Systems for Video Technology , vol. 32, no. 2, pp. 867–879, 2021. 2

  16. [24]

    Joint representation learning and keypoint detection for cross-view geo-localization,

    J. Lin, Z. Zheng, Z. Zhong, Z. Luo, S. Li, Y . Yang, and N. Sebe, “Joint representation learning and keypoint detection for cross-view geo-localization,” IEEE Transactions on Image Processing , vol. 31, pp. 3780–3792, 2022. 2

  17. [25]

    Attention is all you need,

    A. Vaswani, N. Shazeer, N. Parmar, J. Uszkoreit, L. Jones, A. N. Gomez, Ł. Kaiser, and I. Polosukhin, “Attention is all you need,” Advances in Neural Information Processing Systems , vol. 30, 2017. 2

  18. [26]

    Transfg: A cross-view geo-localization of satellite and uavs imagery pipeline using transformer- based feature aggregation and gradient guidance,

    H. Zhao, K. Ren, T. Yue, C. Zhang, and S. Yuan, “Transfg: A cross-view geo-localization of satellite and uavs imagery pipeline using transformer- based feature aggregation and gradient guidance,” IEEE Transactions on Geoscience and Remote Sensing , vol. 62, pp. 1–12, 2024. 2

  19. [27]

    A transformer-based fea- ture segmentation and region alignment method for uav-view geo- localization,

    M. Dai, J. Hu, J. Zhuang, and E. Zheng, “A transformer-based fea- ture segmentation and region alignment method for uav-view geo- localization,” IEEE Transactions on Circuits and Systems for Video Technology, vol. 32, no. 7, pp. 4376–4389, 2021. 2

  20. [28]

    Effovpr: Effective foundation model utilization for visual place recognition,

    I. Tzachor, B. Lerner, M. Levy, M. Green, T. B. Shalev, G. Habib, D. Samuel, N. K. Zailer, O. Shimshi, N. Darshan, and R. Ben- Ari, “Effovpr: Effective foundation model utilization for visual place recognition,” arXiv preprint arXiv:2405.18065 , 2024. 2

  21. [29]

    Anyloc: Towards universal visual place recognition,

    N. Keetha, A. Mishra, J. Karhade, K. M. Jatavallabhula, S. Scherer, M. Krishna, and S. Garg, “Anyloc: Towards universal visual place recognition,” IEEE Robotics and Automation Letters , 2023. 2, 4, 7, 8

  22. [30]

    Lens: Localization enhanced by nerf synthesis,

    A. Moreau, N. Piasco, D. Tsishkou, B. Stanciulescu, and A. de La Fortelle, “Lens: Localization enhanced by nerf synthesis,” in Conference on Robot Learning . PMLR, 2022, pp. 1347–1356. 2, 3

  23. [31]

    Leveraging neural radiance fields for uncertainty-aware visual localization,

    L. Chen, W. Chen, R. Wang, and M. Pollefeys, “Leveraging neural radiance fields for uncertainty-aware visual localization,” in 2024 IEEE International Conference on Robotics and Automation (ICRA) . IEEE, 2024, pp. 6298–6305. 2

  24. [32]

    Uav-satellite view syn- thesis for cross-view geo-localization,

    X. Tian, J. Shao, D. Ouyang, and H. T. Shen, “Uav-satellite view syn- thesis for cross-view geo-localization,” IEEE Transactions on Circuits and Systems for Video Technology, vol. 32, no. 7, pp. 4804–4815, 2022. 3, 8

  25. [33]

    24/7 place recognition by view synthesis,

    A. Torii, R. Arandjelovi ´c, J. Sivic, M. Okutomi, and T. Pajdla, “24/7 place recognition by view synthesis,” in Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition , 2015, pp. 1808–1817. 3

  26. [34]

    Cross-view image synthesis using conditional gans,

    K. Regmi and A. Borji, “Cross-view image synthesis using conditional gans,” in Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition, 2018, pp. 3501–3510. 3

  27. [35]

    Coming down to earth: Satellite-to-street view synthesis for geo-localization,

    A. Toker, Q. Zhou, M. Maximov, and L. Leal-Taixé, “Coming down to earth: Satellite-to-street view synthesis for geo-localization,” in Pro- ceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 2021, pp. 6488–6497. 3

  28. [36]

    Cross-view panorama image synthesis,

    S. Wu, H. Tang, X.-Y . Jing, H. Zhao, J. Qian, N. Sebe, and Y . Yan, “Cross-view panorama image synthesis,” IEEE Transactions on Multi- media, vol. 25, pp. 3546–3559, 2022. 3

  29. [37]

    Denoising diffusion probabilistic models,

    J. Ho, A. Jain, and P. Abbeel, “Denoising diffusion probabilistic models,” Advances in Neural Information Processing Systems , vol. 33, pp. 6840– 6851, 2020. 3

  30. [38]

    Sat2scene: 3d urban scene generation from satellite images with diffusion,

    Z. Li, Z. Li, Z. Cui, M. Pollefeys, and M. R. Oswald, “Sat2scene: 3d urban scene generation from satellite images with diffusion,” in Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, June 2024, pp. 7141–7150. 3

  31. [39]

    Sat2density: Faithful density learning from satellite-ground image pairs,

    M. Qian, J. Xiong, G.-S. Xia, and N. Xue, “Sat2density: Faithful density learning from satellite-ground image pairs,” in Proceedings of the IEEE/CVF International Conference on Computer Vision , 2023, pp. 3683–3692. 3

  32. [40]

    Geometry-guided street- view panorama synthesis from satellite imagery,

    Y . Shi, D. Campbell, X. Yu, and H. Li, “Geometry-guided street- view panorama synthesis from satellite imagery,” IEEE Transactions on Pattern Analysis and Machine Intelligence , vol. 44, no. 12, pp. 10 009– 10 022, 2022. 3

  33. [41]

    Geometry-aware satellite-to-ground image synthesis for urban areas,

    X. Lu, Z. Li, Z. Cui, M. R. Oswald, M. Pollefeys, and R. Qin, “Geometry-aware satellite-to-ground image synthesis for urban areas,” in Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, June 2020. 3

  34. [42]

    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,” Advances in Neural Information Processing Systems , vol. 36, 2024. 3

  35. [43]

    Cross- view meets diffusion: Aerial image synthesis with geometry and text guidance,

    A. Arrabi, X. Zhang, W. Sultan, C. Chen, and S. Wshah, “Cross- view meets diffusion: Aerial image synthesis with geometry and text guidance,” arXiv preprint arXiv:2408.04224 , 2024. 3

  36. [44]

    Hierarchical text-conditional image generation with clip latents,

    A. Ramesh, P. Dhariwal, A. Nichol, C. Chu, and M. Chen, “Hierarchical text-conditional image generation with clip latents,” arXiv preprint arXiv:2204.06125, vol. 1, no. 2, p. 3, 2022. 3

  37. [45]

    High- resolution image synthesis with latent diffusion models,

    R. Rombach, A. Blattmann, D. Lorenz, P. Esser, and B. Ommer, “High- resolution image synthesis with latent diffusion models,” in Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recogni- tion, June 2022, pp. 10 684–10 695. 3

  38. [46]

    From structure- from-motion point clouds to fast location recognition,

    A. Irschara, C. Zach, J.-M. Frahm, and H. Bischof, “From structure- from-motion point clouds to fast location recognition,” in Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 2009, pp. 2599–2606. 3

  39. [47]

    Reference pose generation for long-term visual localization via learned features and view synthesis,

    Z. Zhang, T. Sattler, and D. Scaramuzza, “Reference pose generation for long-term visual localization via learned features and view synthesis,” International Journal of Computer Vision , vol. 129, no. 4, pp. 821–844,

  40. [48]

    Structure-from-motion revisited,

    J. L. Schonberger and J.-M. Frahm, “Structure-from-motion revisited,” in Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition, 2016, pp. 4104–4113. 3, 6

  41. [49]

    Global Structure-from-Motion Revisited,

    L. Pan, D. Barath, M. Pollefeys, and J. L. Schönberger, “Global Structure-from-Motion Revisited,” in Proceedings of the European Con- ference on Computer Vision , 2024. 3

  42. [50]

    The nerfect match: Exploring nerf features for visual localization,

    Q. Zhou, M. Maximov, O. Litany, and L. Leal-Taixé, “The nerfect match: Exploring nerf features for visual localization,” in Proceedings of the European Conference on Computer Vision , 2024. 3

  43. [51]

    Nerf: Representing scenes as neural radiance fields for view synthesis,

    B. Mildenhall, P. P. Srinivasan, M. Tancik, J. T. Barron, R. Ramamoorthi, and R. Ng, “Nerf: Representing scenes as neural radiance fields for view synthesis,” Communications of the ACM , vol. 65, no. 1, pp. 99–106,

  44. [52]

    3d gaussian splatting for real-time radiance field rendering,

    B. Kerbl, G. Kopanas, T. Leimkühler, and G. Drettakis, “3d gaussian splatting for real-time radiance field rendering,” ACM Transactions on Graphics, vol. 42, no. 4, July 2023. 3

  45. [53]

    3dgs-reloc: 3d gaussian splat- ting for map representation and visual relocalization,

    P. Jiang, G. Pandey, and S. Saripalli, “3dgs-reloc: 3d gaussian splat- ting for map representation and visual relocalization,” arXiv preprint arXiv:2403.11367, 2024. 3

  46. [54]

    Renderable street view map-based localization: Leveraging 3d gaussian splatting for street-level positioning,

    J. Howoong, Y . Hyeonwoo, and O. Songhwai, “Renderable street view map-based localization: Leveraging 3d gaussian splatting for street-level positioning,” in Proceedings of the IEEE/RSJ International Conference on Intelligent Robots and Systems , 2024. 3

  47. [55]

    Random sample consensus: a paradigm for model fitting with applications to image analysis and automated cartography,

    M. A. Fischler and R. C. Bolles, “Random sample consensus: a paradigm for model fitting with applications to image analysis and automated cartography,” Communications of the ACM, vol. 24, no. 6, pp. 381–395,

  48. [56]

    Fine-tuning cnn image retrieval with no human annotation,

    F. Radenovi ´c, G. Tolias, and O. Chum, “Fine-tuning cnn image retrieval with no human annotation,” IEEE Transactions on Pattern Analysis and Machine Intelligence, vol. 41, no. 7, pp. 1655–1668, 2018. 4

Pith tools

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