Pith. sign in

REVIEW 3 major objections 5 minor 1 cited by

Around the World in 80 Timesteps: A Generative Approach to Global Visual Geolocation

T0 review · 3 major / 5 minor · reviewed 2026-08-11 · deepseek-v4-flash

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

Pith's one-line read A generative model locates images by denoising random points on the Earth's surface.

desk verdict First generative geolocation on the sphere with strong results, but the published inference equations and appendix need corrections before the probabilistic claims are reliable. read the letter →

arxiv 2412.06781 v1 pith:AK2GEZCY submitted 2024-12-09 cs.CV cs.LG

Nicolas Dufour , David Picard , Vicky Kalogeiton , Loic Landrieu This is my paper · ORCID
classification cs.CVcs.LG
keywords visualgeolocationgenerativemodeldiffusionmodelsflowmatchingRiemannianmanifoldsprobabilisticpredictionlocalizabilitysphereS^2
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 proposes replacing deterministic visual geolocation with a generative process: start from random points on the globe and iteratively denoise them, conditioned on a frozen image embedding, until they converge on plausible capture locations. Because the process runs on the sphere, the same model yields both a point estimate and a full conditional distribution over every location on Earth. The authors report higher accuracy than prior methods on OpenStreetView-5M, YFCC-100M, and iNat21, and they introduce the task of probabilistic visual geolocation with new metrics and baselines. The reason to care is that ambiguity is intrinsic to geolocation, so a density over locations is a more honest and more useful output than a single point.

What carries the argument

The core object is a conditional denoising network ψ that takes as input a noisy coordinate xt, the noise level κ(t), and a frozen image embedding, and learns to predict the noise (diffusion) or the velocity field (flow matching) that moves points toward the true location. The Riemannian flow matching variant operates entirely on the sphere S2 using exponential and logarithmic maps, so every intermediate point lies on the Earth's surface. The probability-density claim is carried by an ODE system that integrates the negative divergence of the learned velocity field along the trajectory, turning the generative model into a likelihood estimator.

What would settle it

Train the same flow-matching model on a synthetic spherical dataset with a known closed-form density, such as a mixture of von Mises distributions, compute likelihoods both by the paper's ODE formula and by the closed form, and check whether the NLL values match; a sign or normalization error would show up directly in this comparison.

Watch

Extended reading notes

Core claim

The central claim is that visual geolocation can be reframed as a generative modeling problem: a network is trained to reverse a noising process that starts from an image's true location and corrupts it into pure noise, so that at inference time the network denoises a random starting point into a likely location. The authors implement this with three variants, diffusion in R3, flow matching in R3, and Riemannian flow matching directly on the sphere S2, and report that the spherical variant performs best, improving GeoScore by hundreds of points over the previous hybrid approach on OpenStreetView-5M while also giving lower average distance and higher country-level accuracy. Beyond the point prediction, the paper claims the learned velocity field can be integrated with a divergence term to compute log p(y|c) for any location y, which yields a calibrated probability distribution over the globe and a quantification of an image's localizability. The paper introduces probabilistic visual geolocation as a task, proposes NLL, localizability, precision, recall, density, and coverage as metrics, and reports that the diffusion and flow-matching models beat von Mises-Fisher baselines on these metrics.

Load-bearing premise

The whole probabilistic side of the paper rests on a formula saying that solving one differential equation along the model's trajectories gives true likelihoods; the appendix's corrected statement disagrees with its own proof about a sign, and the diffusion velocity formula omits a normalization factor, so if those are not harmless errors the reported distribution numbers collapse.

Editorial extensions

If this is right

  • For any new image, the model can return both a single location estimate and a full probability density, so downstream systems can use the distribution to express confidence or to list multiple plausible locations instead of committing to one point.
  • The guidance scale acts as a dial between localization accuracy and distribution fidelity: higher guidance sharpens the predicted density and improves geolocation scores, while lower guidance preserves the full ambiguity of the image.
  • Because the denoising field is defined on the sphere rather than in Euclidean space, the approach respects the Earth's geometry and avoids the projection step that the Euclidean variants require; the authors report that the spherical variant consistently outperforms the Euclidean ones.
  • On YFCC-100M, extending training from one million to ten million iterations yields steady accuracy gains, suggesting the model continues to improve with more compute rather than saturating quickly.
  • The predicted distributions can distinguish images that are pinpointable to meter-level accuracy, such as the Eiffel Tower, from images with only continental information, such as a featureless beach, which is directly useful for applications like forensic verification and archival prioritization.

Reading between the lines

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

  • Editorial inference: if the density estimates are truly calibrated, the model could serve as a spatial prior that is re-weighted by additional cues such as known country, language, or time of day, a combination the paper does not explore.
  • Editorial inference: the same spherical flow-matching setup should transfer to other spherical regression tasks, such as camera pose estimation or planet-scale remote sensing, where injecting Euclidean noise is geometrically inappropriate; the paper only tests it on geolocation.
  • Editorial inference: the distribution over locations could be used as a proposal for retrieval, sampling multiple trajectories and ranking database images by the computed density, potentially improving the fine-grained accuracy where retrieval methods currently beat generative ones.
  • Editorial inference: the reported NLL and localizability numbers depend on the appendix's density formula, and the appendix contains a sign discrepancy between its corrected proposition and its proof, so an independent check on a synthetic spherical density would be the cheapest way to verify the probabilistic results before building on them.
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 / 5 minor

Summary. The paper proposes a generative approach to global visual geolocation. Instead of predicting a single point, the model denoises random locations conditioned on image embeddings, using Euclidean diffusion, Euclidean flow matching, or Riemannian flow matching on the sphere S^2. The authors claim state-of-the-art accuracy on OpenStreetView-5M, YFCC-100M, and iNat21, and introduce a probabilistic variant in which the model outputs a full distribution over locations via an ODE-based density computation. They also propose new metrics and baselines for probabilistic geolocation and show qualitative localizability results.

Significance. If the claims hold, this is a substantial contribution: it is the first generative treatment of global visual geolocation, it provides calibrated spatial distributions in addition to point predictions, and it demonstrates strong empirical results on three large benchmarks. The paper also adapts flow-matching density estimation to a compact manifold and introduces a new evaluation protocol for probabilistic geolocation, which could be useful for future work. The experimental comparison is broad, and the inclusion of non-parametric generative baselines (vMF, vMF mixtures) is appropriate.

major comments (3)
  1. [Sec. 3.1, Eqs. (3) and (4)] The sampling procedure as printed is an identity. Substituting Eq. (4) into Eq. (3) gives x_{t-dt} = sqrt(1-kappa(t)) * [1/sqrt(1-kappa(t)) (x_t - sqrt(kappa(t)) psi(x_t|c))] + sqrt(kappa(t)) psi(x_t|c) = x_t, identically for any network output. The iterative denoising loop therefore does not modify the initial noise, and after projection the result would be a uniform random point on S^2, which cannot produce the accuracies reported in Table 1. The likely intended update uses kappa(t-dt) in Eq. (3) (standard DDIM indexing), but the manuscript does not state this and no code is available for verification. This is a load-bearing defect in the central inference algorithm and must be corrected and re-verified.
  2. [Appendix D, Proposition 2 and Eq. (W)] The 'corrected' Proposition 2 contains a sign error. Its statement claims log p(y|c) = log p_epsilon(x(1)|c) + f(1), while the proof in the same appendix (Eq. (O)) derives log p(x_0|c) = log p(x(1)|c) - f(1), which is consistent with Proposition 1 in the main text. The two are mutually contradictory. Additionally, Eq. (W) omits the 1/sqrt(kappa(t)) normalization when substituting the learned noise for the score: the correct velocity for the probability-flow ODE is -1/2 beta(t) (x - eps_theta / sqrt(kappa(t))), not -1/2 beta(t) (x - eps_theta). This affects the diffusion-model density estimates, so the DiffR3 NLL values and localizability numbers in Table 2 are unreliable as printed.
  3. [Sec. 4.2 / Table 2] The NLL comparison in Table 2 mixes models defined in different metric spaces (R^3 versus S^2). The paper notes that these are not directly comparable, but then relies on this table to support the probabilistic-geolocation claim. Since the NLL of the R^3 models depends on an unnormalized density in the embedding space, the reported numerical differences (e.g., Diffusion R^3 NLL 0.58 vs. RFlowMatch S^2 NLL -1.51 on OSV-5M) should not be interpreted as relative quality without a common normalization; at minimum, the R^3 densities must be marginalized onto the sphere consistently before comparison.
minor comments (5)
  1. [Table A / Sec. A] The ablation text says the standard sigmoid scheduler 'increases the geoscore' relative to the proposed scheduler, and the table confirms this (3767.21 vs. 3746.79), yet the main text and Fig. 4 argue that the skewed scheduler gives better results. This contradiction should be clarified: either the scheduler choice is justified by density quality alone, or the numbers are misreported.
  2. [Abstract / Section 3.4] The abstract states 'Codes and models are available here' but no URL is given in the manuscript. Please provide a working link or state that code will be released upon publication.
  3. [Table 1] The row for 'Uniform' under OSV-5M contains an excessive number of numerical entries (six values for five columns), making the table hard to read; please fix the alignment.
  4. [Sec. 4.2] There is a typo: 'New Zeland' should be 'New Zealand'.
  5. [Appendix A] The phrase 'as measurs by the generative metrics' contains a typo; it should read 'as measured by the generative metrics'.

Circularity Check

0 steps flagged · score 0.0 of 10

No circularity: the derivation is a standard flow-matching pipeline and the evaluation is against fixed external benchmarks; the printed DDIM and density-formula inconsistencies are correctness defects, not circular reasoning.

full rationale

We find no circular step in the paper's derivation chain. The core training procedure fits a network psi to regress either noise (Eq. 2) or a velocity field (Eqs. 7, 11), and inference integrates that learned field from independently sampled noise (Eqs. 3-4, 8, 12); the output is a function of the trained network and a random initial sample, not a re-insertion of the fitted quantity as its own prediction. The density estimate in Proposition 1 / Proposition 2 is the standard flow-matching change-of-variables identity derived from the logarithmic mass conservation theorem, with p_epsilon as the known prior noise distribution; it does not assume the target density p(y|c) it claims to compute. The self-citations to OpenStreetView-5M [2] and to earlier diffusion papers from the same group provide a dataset and background context, but the state-of-the-art claims are evaluated on fixed public test sets with external baselines, so the self-citations are not load-bearing. The algebraic identity in the printed DDIM update equations (Eqs. 3 and 4) and the sign inconsistency between the appendix's corrected Proposition 2 and its proof are serious correctness/consistency defects that would invalidate the reported numbers as printed, but they are not circularity: they do not make a claimed result equivalent to its input by construction. The scheduler and guidance scale are tuned on validation data, which is standard practice and does not constitute fitted-input-called-prediction. Overall, the paper's argument is self-contained and non-circular, though it contains internal mathematical errors that require correction.

Assumptions & free parameters 5 free parameters · 6 assumptions · 0 invented entities

The central geolocation result rests on standard deep-learning assumptions (good encoder, fair benchmarks) and on the Riemannian flow matching framework. The probabilistic result additionally rests on the divergence-based density formula, which contains the sign and normalization errors noted in red_flags.

free parameters (5)
  • Scheduler skewness (alpha, beta) = alpha=-3, beta=7 in text; beta=3 in ablation
    Chosen by validation; the text/ablation conflict makes the effective value unclear.
  • Guidance scale omega = 2 (geolocation), 0 (density)
    Tuned on OSV-5M validation to balance GeoScore and distribution quality.
  • Number of sampling timesteps = 16-32
    GeoScore plateaus around 16 steps (Fig 5), so additional steps are a compute-accuracy trade-off.
  • k for precision/recall/kNN metrics = k=3
    Selected for the adapted manifold metrics; affects the reported density/coverage numbers.
  • vMF mixture components K (baseline) = K=3
    Baseline configuration, not part of the main method.
assumptions (6)
  • domain assumption The frozen image encoder (DINOv2-L or StreetCLIP ViT-L) provides a sufficiently rich visual representation for location prediction.
    The whole method conditions on this embedding; if it lacks location cues, the small denoising network cannot recover them. Invoked in Sec 3.4.
  • domain assumption The Earth's surface is adequately modeled as the unit sphere S2.
    Riemannian flow matching requires a manifold model; the oblateness of the Earth is ignored, which is standard in geolocation literature.
  • standard math The logarithmic mass conservation theorem governs the probability density evolution for the learned velocity field.
    Used in Proposition 1/2 to relate divergence to log-density; a standard result from optimal transport.
  • standard math Riemannian flow matching on S2 with exponential/log maps as given is a valid generative framework.
    The paper builds on Chen and Lipman (ICLR 2024) and uses the maps in Appendix D.
  • domain assumption Benchmark test sets are free of train/test leakage (OSV-5M has a 1km buffer; iNat21 and YFCC4k rely on the original protocols).
    SOTA claims depend on fair evaluation; the paper does not perform its own leakage analysis.
  • domain assumption The conditional distribution p(y|c) is well represented as the pushforward of the noise distribution under the learned ODE.
    Needed for the density-estimation step; a standard but nontrivial modeling assumption for flow matching.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Around the World in 80 Timesteps: A Generative Approach to Global Visual Geolocation." pith.science (2026). https://pith.science/paper/AK2GEZCY

@misc{pith2026241206781,
  author       = {Pith},
  title        = {Pith review of: Around the World in 80 Timesteps: A Generative Approach to Global Visual Geolocation},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/AK2GEZCY}},
  note         = {Machine review of arXiv:2412.06781}
}
read the original abstract

Global visual geolocation predicts where an image was captured on Earth. Since images vary in how precisely they can be localized, this task inherently involves a significant degree of ambiguity. However, existing approaches are deterministic and overlook this aspect. In this paper, we aim to close the gap between traditional geolocalization and modern generative methods. We propose the first generative geolocation approach based on diffusion and Riemannian flow matching, where the denoising process operates directly on the Earth's surface. Our model achieves state-of-the-art performance on three visual geolocation benchmarks: OpenStreetView-5M, YFCC-100M, and iNat21. In addition, we introduce the task of probabilistic visual geolocation, where the model predicts a probability distribution over all possible locations instead of a single point. We introduce new metrics and baselines for this task, demonstrating the advantages of our diffusion-based approach. Codes and models will be made available.

Figures

Figures reproduced from arXiv: 2412.06781 by the authors.

Figure 1
Figure 1. Geolocation as a Generative Process. We explore diffusion and flow matching for visual geolocation by sampling and denoising random locations. This process generates trajectories onto the Earth’s surface, whose endpoints provide location estimates. Our models also provide probability densities for every possible image locations. We illustrate these trajectories and the log-densities for three images from different d… view at source ↗
Figure 2
Figure 2. Generative Framework. We implement three generative approaches for geolocation: diffusion in R 3 , flow matching in R 3 , and Riemannian flow matching directly on S2. This figure provides the formulas for the noising processes and the loss functions for each approach. on κ(t) in the rest of the paper. The model is trained to minimize the diffusion loss function: \label {eq:loss_diff} \mathcal {L}_{\text {D}} = \math… view at source ↗
Figure 3
Figure 3. Inference Pipeline. We start by embedding the image to be localized into a vector using a frozen image encoder. We then sample a random noise ϵ in R 3 or on S2, projected here onto the sphere. We iteratively remove the noise using either the reverse diffusion or flow-matching equations for t = 1 to 0. The final point of this trajectory is our predicted location. Additionally, our model be queried to predict a probab… view at source ↗
Figures from the paper (4 more)
Figure 4
Figure 4. Figure 4: Scheduler. We chose a noise scheduler that assigns more weights to the beginning of the diffusion process. Guided Geolocation. We adapt the idea of classifier-free guidance [34] to our setting. We train the network ψ to learn both the conditional distribution p(y | c) …
Figure 5
Figure 5. Figure 5: Impact of Number of Timesteps. We represent different metrics on OpenStreetView-5M with different numbers of timesteps for the Riemannian Flow matching model. K vMF distributions (vMFMix) with mixture weights w1, . . . , wK > 0 and PK k=1 wk = 1, and distribution pa￾ra…
Figure 6
Figure 6. Figure 6: Estimating Localizability. We use the entropy of the predicted distribution as a proxy for the localizability of images. For each dataset, we present examples of high, medium, and low localizability, which correlate well with human perception. sphere S 2 due to differe…
Figure 7
Figure 7. Figure 7: Impact of Classifier-Free Guidance. We plot the evo￾lution of the GeoScore and generative metrics depending on the guidance scale ω for the OSV-5M dataset. generation process, compared to R 3 where the output of the generation process has to be projected onto S 2 which…

Discussion (0). Continue with ORCID to comment.

Forward citations

Cited by 1 Pith paper

Reviewed papers in the Pith corpus that reference this work. Sorted by Pith novelty score. Full citation record

  1. GeoLocSFT: Efficient Visual Geolocation via Supervised Fine-Tuning of Multimodal Foundation Models

    cs.AI 2025-06 conditional novelty 6.0 of 10

    Fine-tuning Gemma 3 on 2,700 LLM-generated geo-captions gives competitive image geolocation and a new MR40k rural benchmark.

Reference graph

Works this paper leans on

82 extracted references · 74 canonical work pages · cited by 1 Pith paper

  1. [1]

    https : / / gitlab

    YFCC100m. https : / / gitlab . com / jfolz / yfcc100m, accessed: 2023-10-10 1, 2, 6, 8

    2023

  2. [2]

    In: CVPR (2024) 1, 2, 5, 6, 7, 8

    Astruc, G., Dufour, N., Siglidis, I., Aronssohn, C., Bouia, N., Fu, S., Loiseau, R., Nguyen, V .N., Raude, C., Vincent, E., et al.: OpenStreetView-5M: The many roads to global visual geolocation. In: CVPR (2024) 1, 2, 5, 6, 7, 8

    2024

  3. [3]

    In: arXiv preprint arXiv:2402.15448 (2024) 1

    Bamigbade, O., Sheppard, J., Scanlon, M.: Computer vision for multimedia geolocation in human trafficking investigation: A systematic literature review. In: arXiv preprint arXiv:2402.15448 (2024) 1

    arXiv 2024

  4. [4]

    In: ICML (2022) 5, 15

    Ben-Hamu, H., Cohen, S., Bose, J., Amos, B., Nickel, M., Grover, A., Chen, R.T., Lipman, Y .: Matching normalizing flows and probability paths on manifolds. In: ICML (2022) 5, 15

    work page 2022

  5. [5]

    In: UAI (2024) 2

    Berry, L., Brando, A., Meger, D.: Shedding light on large generative networks: Estimating epistemic uncer- tainty in diffusion models. In: UAI (2024) 2

    work page 2024

  6. [6]

    arXiv preprint arXiv:2311.15127 (2023) 2

    Blattmann, A., Dockhorn, T., Kulal, S., Mendelevitch, D., Kilian, M., Lorenz, D., Levi, Y ., English, Z., V oleti, V ., Letts, A., et al.: Stable video diffusion: Scaling latent video diffusion models to large datasets. arXiv preprint arXiv:2311.15127 (2023) 2

    arXiv 2023

  7. [7]

    Numerical Analysis and Optimization (2015) 5

    Butcher, J.: Runge-Kutta methods for ordinary differ- ential equations. Numerical Analysis and Optimization (2015) 5

    work page 2015

  8. [8]

    Chen, R.T.Q.: torchdiffeq (2018), https : / / github.com/rtqichen/torchdiffeq 5

    work page 2018

Show all 82 references
  1. [9]

    In: ICLR (2024) 2, 4, 7

    Chen, R.T., Lipman, Y .: Riemannian flow matching on general geometries. In: ICLR (2024) 2, 4, 7

    2024

  2. [10]

    In: ICLR (2024) 2

    Chen, R.T., Lipman, Y .: Riemannian flow matching on general geometries. In: ICLR (2024) 2

    2024

  3. [11]

    In: CVPR (2023) 2, 6

    Clark, B., Kerrigan, A., Kulkarni, P.P., Cepeda, V .V ., Shah, M.: Where we are and what we’re looking at: Query based worldwide image geo-localization using hierarchies and scenes. In: CVPR (2023) 2, 6

    2023

  4. [12]

    In: ECCV (2024) 2

    Courant, R., Dufour, N., Wang, X., Christie, M., Kalo- geiton, V .: ET the exceptional trajectories: Text-to- camera-trajectory generation with character awareness. In: ECCV (2024) 2

    2024

  5. [13]

    International Journal of Geographical Information Science (2013) 1

    Daoud, M., Huang, J.X.: Mining query-driven con- texts for geographic and temporal search. International Journal of Geographical Information Science (2013) 1

    2013

  6. [14]

    ICLR (2024) 6

    Darcet, T., Oquab, M., Mairal, J., Bojanowski, P.: Vi- sion transformers need registers. ICLR (2024) 6

    2024

  7. [15]

    In: ICRA (1999) 2

    Dellaert, F., Fox, D., Burgard, W., Thrun, S.: Monte Carlo localization for mobile robots. In: ICRA (1999) 2

    1999

  8. [16]

    In: ACL Linguistic Annotation Workshop (2016) 1

    DeLozier, G., Wing, B., Baldridge, J., Nesbit, S.: Cre- ating a novel geolocation corpus from historical texts. In: ACL Linguistic Annotation Workshop (2016) 1

    2016

  9. [17]

    Interna- tional Journal of Computer Vision (2022) 2

    Deng, H., Bui, M., Navab, N., Guibas, L., Ilic, S., Birdal, T.: Deep Bingham networks: Dealing with uncertainty and ambiguity in pose estimation. Interna- tional Journal of Computer Vision (2022) 2

    2022

  10. [18]

    Journal of computational and applied mathematics (1980) 5

    Dormand, J.R., Prince, P.J.: A family of embedded Runge-Kutta formulae. Journal of computational and applied mathematics (1980) 5

    1980

  11. [19]

    ICLR (2021) 6

    Dosovitskiy, A.: An image is worth 16x16 words: Transformers for image recognition at scale. ICLR (2021) 6

    2021

  12. [20]

    In: CVPR (2024) 2

    Dufour, N., Besnier, V ., Kalogeiton, V ., Picard, D.: Don’t drop your samples! Coherence-aware training benefits conditional diffusion. In: CVPR (2024) 2

    2024

  13. [21]

    Cambridge university press (2019) 15

    Durrett, R., Durrett, R.: Probability: Theory and exam- ples. Cambridge university press (2019) 15

    2019

  14. [22]

    Proceedings of the Royal Society of London

    Fisher, R.A.: Dispersion on a sphere. Proceedings of the Royal Society of London. Series A. Mathematical and Physical Sciences (1953) 6

    1953

  15. [23]

    In: International Conference on Web Search and Data Mining (2015) 1

    Flatow, D., Naaman, M., Xie, K.E., V olkovich, Y ., Kanza, Y .: On the accuracy of hyper-local geotagging of social media content. In: International Conference on Web Search and Data Mining (2015) 1

    2015

  16. [24]

    In: ICLR (2019) 2

    Grathwohl, W., Chen, R.T., Bettencourt, J., Sutskever, I., Duvenaud, D.: FFJORD: Free-form continuous dy- namics for scalable reversible generative models. In: ICLR (2019) 2

    2019

  17. [25]

    In: arXiv preprint arXiv:2302.00275 (2023) 6

    Haas, L., Alberti, S., Skreta, M.: Learning generalized zero-shot learners for open-domain image geolocal- ization. In: arXiv preprint arXiv:2302.00275 (2023) 6

    2023 arXiv

  18. [26]

    In: CVPR (2023) 2, 6

    Haas, L., Alberti, S., Skreta, M.: PIGEON: Predicting image geolocations. In: CVPR (2023) 2, 6

    2023

  19. [27]

    In: arXiv preprint arXiv:1706.04264 (2017) 6

    Hasnat, M.A., Bohn´e, J., Milgram, J., Gentric, S., Chen, L.: von Mises-Fisher mixture model-based deep learn- ing: Application to face verification. In: arXiv preprint arXiv:1706.04264 (2017) 6

    2017 arXiv

  20. [28]

    In: CVPR (2008) 1, 2

    Hays, J., Efros, A.A.: Im2GPSs: Estimating geo- graphic information from a single image. In: CVPR (2008) 1, 2

    2008

  21. [29]

    Multimodal location estimation of videos and images (2015) 2

    Hays, J., Efros, A.A.: Large-scale image geolocaliza- tion. Multimodal location estimation of videos and images (2015) 2

    2015

  22. [30]

    In: arXiv preprint arXiv:1606.08415 (2016) 5

    Hendrycks, D., Gimpel, K.: Gaussian error linear units (Gelus). In: arXiv preprint arXiv:1606.08415 (2016) 5

    2016 arXiv

  23. [31]

    arXiv (2022) 2

    Ho, J., Chan, W., Saharia, C., Whang, J., Gao, R., Gritsenko, A., Kingma, D.P., Poole, B., Norouzi, M., Fleet, D.J., et al.: Imagen video: High definition video generation with diffusion models. arXiv (2022) 2

    2022

  24. [32]

    In: NeurIPS (2020) 2, 3

    Ho, J., Jain, A., Abbeel, P.: Denoising diffusion proba- bilistic models. In: NeurIPS (2020) 2, 3

    2020

  25. [33]

    Journal of Machine Learning Research (2022) 2

    Ho, J., Saharia, C., Chan, W., Fleet, D.J., Norouzi, M., Salimans, T.: Cascaded diffusion models for high fidelity image generation. Journal of Machine Learning Research (2022) 2

    2022

  26. [34]

    In: NeurIPS 2021 Workshop on Deep Generative Mod- els and Downstream Applications (2021) 5

    Ho, J., Salimans, T.: Classifier-free diffusion guidance. In: NeurIPS 2021 Workshop on Deep Generative Mod- els and Downstream Applications (2021) 5

    2021

  27. [35]

    In: ECML-PKDD

    Huang, B., Yu, W., Xie, R., Xiao, J., Huang, J.: Two- stage denoising diffusion model for source localization in graph inverse problems. In: ECML-PKDD. Springer (2023) 2

    2023

  28. [36]

    In: MLKDD (2020) 2, 6

    Izbicki, M., Papalexakis, E.E., Tsotras, V .J.: Exploiting the Earth’s spherical geometry to geolocate images. In: MLKDD (2020) 2, 6

    2020

  29. [37]

    In: ICRA (2016) 2

    Kendall, A., Cipolla, R.: Modelling uncertainty in deep learning for camera relocalization. In: ICRA (2016) 2

    2016

  30. [38]

    Kendall, A., Gal, Y .: What uncertainties do we need in bayesian deep learning for computer vision? In: NeurIPS (2017) 2

    2017

  31. [39]

    In: International Conference on Multimedia Retrieval (2021) 2

    Kordopatis-Zilos, G., Galopoulos, P., Papadopoulos, S., Kompatsiaris, I.: Leveraging EfficientNet and con- trastive learning for accurate global-scale location es- timation. In: International Conference on Multimedia Retrieval (2021) 2

    2021

  32. [40]

    NeurIPS (2019) 7, 14

    Kynk¨a¨anniemi, T., Karras, T., Laine, S., Lehtinen, J., Aila, T.: Improved precision and recall metric for as- sessing generative models. NeurIPS (2019) 7, 14

    2019

  33. [41]

    In: ICRA (2010) 2

    Levinson, J., Thrun, S.: Robust vehicle localization in urban environments using probabilistic maps. In: ICRA (2010) 2

    2010

  34. [42]

    In: CVPR (2023) 3

    Li, A.C., Prabhudesai, M., Duggal, S., Brown, E., Pathak, D.: Your diffusion model is secretly a zero- shot classifier. In: CVPR (2023) 3

    2023

  35. [43]

    In: CVPR (2024) 2

    Li, W., Yang, Y ., Yu, S., Hu, G., Wen, C., Cheng, M., Wang, C.: Diffloc: Diffusion model for outdoor lidar localization. In: CVPR (2024) 2

    2024

  36. [44]

    In: ICLR (2023) 2, 15

    Lipman, Y ., Chen, R.T., Ben-Hamu, H., Nickel, M., Le, M.: Flow matching for generative modeling. In: ICLR (2023) 2, 15

    2023

  37. [45]

    In: The Eleventh International Conference on Learning Representations (2024) 2, 3, 5

    Lipman, Y ., Chen, R.T., Ben-Hamu, H., Nickel, M., Le, M.: Flow matching for generative modeling. In: The Eleventh International Conference on Learning Representations (2024) 2, 3, 5

    2024

  38. [46]

    In: CVPR (2021) 2

    Mackowiak, R., Ardizzone, L., Kothe, U., Rother, C.: Generative classifiers as a basis for trustworthy image classification. In: CVPR (2021) 2

    2021

  39. [47]

    In: ICCV (2001) 2

    Martin, D., Fowlkes, C., Tal, D., Malik, J.: A database of human segmented natural images and its application to evaluating segmentation algorithms and measuring ecological statistics. In: ICCV (2001) 2

    2001

  40. [48]

    In: CVPR (2022) 2

    Merrill, N., Guo, Y ., Zuo, X., Huang, X., Leuteneg- ger, S., Peng, X., Ren, L., Huang, G.: Symmetry and uncertainty-aware object SLAM for 6DOF object pose estimation. In: CVPR (2022) 2

    2022

  41. [49]

    Journal of the American statistical association (1949) 7

    Metropolis, N., Ulam, S.: The Monte Carlo method. Journal of the American statistical association (1949) 7

    1949

  42. [50]

    ISMIR (2021) 2

    Mittal, G., Engel, J., Hawthorne, C., Simon, I.: Sym- bolic music generation with diffusion models. ISMIR (2021) 2

    2021

  43. [51]

    IEEE transactions on robotics (2011) 2

    Mullane, J., V o, B.N., Adams, M.D., V o, B.T.: A random-finite-set approach to Bayesian SLAM. IEEE transactions on robotics (2011) 2

    2011

  44. [52]

    In: ECCV (2018) 6

    Muller-Budack, E., Pustu-Iren, K., Ewerth, R.: Geolo- cation estimation of photos using a hierarchical model and scene classification. In: ECCV (2018) 6

    2018

  45. [53]

    In: ICML (2020) 7, 14

    Naeem, M.F., Oh, S.J., Uh, Y ., Choi, Y ., Yoo, J.: Reli- able fidelity and diversity metrics for generative models. In: ICML (2020) 7, 14

    2020

  46. [54]

    In: ECCV (2022) 2

    Nicolas Dufour, David Picard, V .K.: SCAM! Trans- ferring humans between images with semantic cross attention modulation. In: ECCV (2022) 2

    2022

  47. [55]

    Inter- national Journal on Document Analysis and Recogni- tion (2022) 1

    Nikolaidou, K., Seuret, M., Mokayed, H., Liwicki, M.: A survey of historical document image datasets. Inter- national Journal on Document Analysis and Recogni- tion (2022) 1

    2022

  48. [56]

    Progress in brain research (2006) 2

    Oliva, A., Torralba, A.: Building the gist of a scene: The role of global image features in recognition. Progress in brain research (2006) 2

    2006

  49. [57]

    Oquab, M., Darcet, T., Moutakanni, T., V o, H.V ., Szafraniec, M., Khalidov, V ., Fernandez, P., HAZIZA, D., Massa, F., El-Nouby, A., et al.: DINOv2: Learning robust visual features without supervision. TMLR 6

  50. [58]

    In: CVPR (2017) 2

    Pavlakos, G., Zhou, X., Derpanis, K.G., Daniilidis, K.: Coarse-to-fine volumetric prediction for single-image 3D human pose. In: CVPR (2017) 2

    2017

  51. [59]

    In: ICCV (2023) 5

    Peebles, W., Xie, S.: Scalable diffusion models with transformers. In: ICCV (2023) 5

    2023

  52. [60]

    In: CVPR Workshop on Human Motion Generation (2024) 2

    Petrovich, M., Litany, O., Iqbal, U., Black, M.J., Varol, G., Peng, X.B., Rempe, D.: Multi-track timeline con- trol for text-driven 3D human motion generation. In: CVPR Workshop on Human Motion Generation (2024) 2

    2024

  53. [61]

    arXiv (2024) 2

    Polyak, A., Zohar, A., Brown, A., Tjandra, A., Sinha, A., Lee, A., Vyas, A., Shi, B., Ma, C.Y ., Chuang, C.Y ., et al.: Movie Gen: A cast of media foundation models. arXiv (2024) 2

    2024

  54. [62]

    In: ICML

    Popov, V ., V ovk, I., Gogoryan, V ., Sadekova, T., Kudi- nov, M.: Grad-TTS: A diffusion probabilistic model for text-to-speech. In: ICML. PMLR (2021) 2

    2021

  55. [63]

    In: ECCV (2022) 2, 6

    Pramanick, S., Nowara, E.M., Gleason, J., Castillo, C.D., Chellappa, R.: Where in the world is this im- age? Transformer-based geo-localization in the wild. In: ECCV (2022) 2, 6

    2022

  56. [64]

    In: CVPR (2022) 2

    Rombach, R., Blattmann, A., Lorenz, D., Esser, P., Om- mer, B.: High-resolution image synthesis with latent diffusion models. In: CVPR (2022) 2

    2022

  57. [65]

    In: NeurIPS (2022) 2

    Saharia, C., Chan, W., Saxena, S., Li, L., Whang, J., Denton, E.L., Ghasemipour, K., Gontijo Lopes, R., Karagol Ayan, B., Salimans, T., et al.: Photorealis- tic text-to-image diffusion models with deep language understanding. In: NeurIPS (2022) 2

    2022

  58. [66]

    In: ECCV (2018) 6

    Seo, P.H., Weyand, T., Sim, J., Han, B.: Cplanet: En- hancing image geolocalization by combinatorial parti- tioning of maps. In: ECCV (2018) 6

    2018

  59. [67]

    In: International Conference on Theory and Practice of Digital Libraries

    Smith, D.A., Crane, G.: Disambiguating geographic names in a historical digital library. In: International Conference on Theory and Practice of Digital Libraries. Springer Berlin Heidelberg, Berlin, Heidelberg (2001) 1

    2001

  60. [68]

    In: Riemannian Geometric Statistics in Medical Image Analysis

    Sommer, S., Fletcher, T., Pennec, X.: Introduction to differential and riemannian geometry. In: Riemannian Geometric Statistics in Medical Image Analysis. Else- vier (2020) 15

    2020

  61. [69]

    In: ICLR (2021) 2

    Song, J., Meng, C., Ermon, S.: Denoising diffusion implicit models. In: ICLR (2021) 2

    2021

  62. [70]

    In: ICLR (2021) 3, 15

    Song, J., Meng, C., Ermon, S.: Denoising diffusion implicit models. In: ICLR (2021) 3, 15

    2021

  63. [71]

    In: ICLR (2021) 2, 3, 15

    Song, Y ., Sohl-Dickstein, J., Kingma, D.P., Kumar, A., Ermon, S., Poole, B.: Score-based generative model- ing through stochastic differential equations. In: ICLR (2021) 2, 3, 15

    2021

  64. [72]

    In: W ACV (2022) 2

    Theiner, J., M ¨uller-Budack, E., Ewerth, R.: Inter- pretable semantic photo geolocation. In: W ACV (2022) 2

    2022

  65. [73]

    In: NeurIPS (2014) 2

    Tompson, J.J., Jain, A., LeCun, Y ., Bregler, C.: Joint training of a convolutional network and a graphical model for human pose estimation. In: NeurIPS (2014) 2

    2014

  66. [74]

    In: CVPR (2021) 1, 2, 5, 6, 8

    Van Horn, G., Cole, E., Beery, S., Wilber, K., Be- longie, S., Mac Aodha, O.: Benchmarking representa- tion learning for natural world image collections. In: CVPR (2021) 1, 2, 5, 6, 8

    2021

  67. [75]

    Berlin: Springer (2009) 5, 15

    Villani, C.: Optimal transport: Old and new. Berlin: Springer (2009) 5, 15

    2009

  68. [76]

    In: ICCV (2017) 1, 2, 6

    V o, N., Jacobs, N., Hays, J.: Revisiting IMG2GPS in the deep learning era. In: ICCV (2017) 1, 2, 6

    2017

  69. [77]

    In: ECCV (2016) 2, 6

    Weyand, T., Kostrikov, I., Philbin, J.: Planet-photo geolocation with convolutional neural networks. In: ECCV (2016) 2, 6

    2016

  70. [78]

    In: International Conference on Medical Imaging with Deep Learning

    Wolleb, J., Sandk¨uhler, R., Bieder, F., Valmaggia, P., Cattin, P.C.: Diffusion models for implicit image seg- mentation ensembles. In: International Conference on Medical Imaging with Deep Learning. PMLR (2022) 2

    2022

  71. [79]

    In: CVPR (2024) 2

    Xu, L., Qu, H., Cai, Y ., Liu, J.: 6D-diff: A keypoint diffusion framework for 6d object pose estimation. In: CVPR (2024) 2

    2024

  72. [80]

    Forensic Science International: Digital Investigation (2020) 1

    Yokota, R., Hawai, Y ., Tsuchiya, K., Imoto, D., Hirabayashi, M., Akiba, N., Kakuda, H., Tanabe, K., Honma, M., Kurosawa, K.: A revisited visual-based geolocalization framework for forensic investigation support tools. Forensic Science International: Digital Investigation (2020) 1

    2020

  73. [81]

    ICLR (2020) 13

    You, Y ., Li, J., Reddi, S., Hseu, J., Kumar, S., Bhojana- palli, S., Song, X., Demmel, J., Keutzer, K., Hsieh, C.J.: Large batch optimization for deep learning: Training bert in 76 minutes. ICLR (2020) 13

    2020

  74. [82]

    In: ICRA (2023) 2 Around the World in 80 Timesteps: A Generative Approach to Global Visual Geolocation Supplementary Material In this appendix, we present our ablation study in Sec

    Zangeneh, F., Bruns, L., Dekel, A., Pieropan, A., Jens- felt, P.: A probabilistic framework for visual localiza- tion in ambiguous scenes. In: ICRA (2023) 2 Around the World in 80 Timesteps: A Generative Approach to Global Visual Geolocation Supplementary Material In this appe...

    2023

Pith tools

Reviewed August 11, 2026 · model on record in the stance chip above.