Pith. sign in

REVIEW 4 major objections 6 minor 49 references

Retrieval-guided Cross-view Image Synthesis

T0 review · 4 major / 6 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 The paper claims that conditioning a GAN on embeddings from a frozen cross-view retrieval network, plus an identity loss matching location identity, achieves state-of-the-art cross-view image synthesis without semantic maps or…

desk verdict A practical retrieval-guided GAN for cross-view synthesis with a new urban dataset, but the retrieval-accuracy claims need re-examination because the identity loss trains against the same embedding family used for evaluation. read the letter →

arxiv 2411.19510 v2 pith:YWHRYUCW submitted 2024-11-29 cs.CV cs.LG

Hongji Yang , Yiru Li , Yingying Zhu This is my paper · ORCID
classification cs.CVcs.LG
keywords cross-viewimagesynthesisretrieval-guidedgenerationgeo-localizationgenerativeadversarialnetworkattentionalAdaINidentitylossurbandatasetsemanticsimilarity
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 a frozen, contrastively trained cross-view retrieval network can replace the semantic segmentation maps and polar-transformation preprocessing that earlier systems use to bridge the aerial-ground domain gap. The authors build a GAN in which the retrieval embedding of the source image conditions both the generator and the discriminator, and an identity loss pushes the embedding of the generated image toward the embedding of the target. On CVUSA, CVACT, and a new urban dataset called VIGOR-GEN it reports lower FID and higher retrieval accuracy (R@1) than prior methods, for example FID 13.57 versus 20.63 for the CDE baseline on CVUSA aerial-to-ground synthesis. If the result holds, retrieval embeddings alone carry enough view-invariant and view-specific signal to synthesize high-fidelity cross-view images without auxiliary inputs.

What carries the argument

The load-bearing machinery is a frozen retrieval embedder--the shallower SAIG-S model from the authors' own retrieval work (reference [46])--that maps aerial and ground images into a shared embedding space trained by contrastive learning. This embedder does three jobs: it supplies the condition vector injected into the generator's residual blocks through attentional AdaIN; it conditions the discriminator so fake images that do not match the source location score low; and it defines the identity loss, $L_{\mathrm{id}} = 1 - \cos(R(X_r), R(X_t)) + 1 - \cos(R(X'_r), R(X'_t))$, which pushes generated and target images toward the same retrieval embedding. The attentional AdaIN module is the paper's fusion mechanism: it normalizes feature maps, modulates them with MLP-predicted scale and shift derived from the retrieval embedding, learns a spatial weight map from the input features, and blends modulated and unmodulated features so the embedding is applied mainly where the two views correspond.

What would settle it

Swap the SAIG-S embedder for one trained with a different objective that is known to produce a non-smooth embedding space (the paper's LPN comparison is a first step), and then, with the generator frozen, re-measure R@1 using a retrieval model trained independently on a different backbone; if the R@1 advantage over baselines collapses or the identity loss stops converging while pixel-level losses are unchanged, the smoothness claim is falsified. A complementary observable test is to inspect the learned spatial mask M in attentional AdaIN: if it does not concentrate on view-invariant structures such as roads, layouts, and building footprints on VIGOR-GEN, the embedding is not guiding correspondence in the way the paper claims.

Watch

Extended reading notes

Core claim

The central discovery is that the embedding space of a contrastive cross-view retrieval model is itself a usable condition for synthesis: contrastive training pulls images of the same place from different viewpoints together, so the embedding encodes view-invariant location identity, and the paper claims this space is smooth enough that the gradients of a cosine-similarity identity loss point the generator toward the correct target appearance. The method injects the fixed retrieval embedding into a GAN generator through attentional AdaIN layers that learn a spatial mask, adds a modulated style path for view-specific detail, and uses a discriminator conditioned on the same embedding. The authors report state-of-the-art SSIM, PSNR, LPIPS, FID, and R@1 numbers on CVUSA and CVACT, and on VIGOR-GEN, a 103,516-pair urban dataset they introduce, they likewise report the best scores, including FID 13.64 for aerial-to-ground synthesis.

Load-bearing premise

The load-bearing premise is that the fixed retrieval network has a smooth, semantically grounded embedding space, so the cosine-identity loss gives reliable gradients for view-invariant content; if that space is noisy, easily fooled, or mostly encodes view-specific appearance, the model will overfit to the embedder and the reported retrieval and realism gains will not transfer.

Editorial extensions

If this is right

  • Semantic segmentation maps and polar-transformation preprocessing become unnecessary for cross-view synthesis, and the same frozen embedder serves both aerial-to-ground and ground-to-aerial generation.
  • Synthesis and retrieval objectives no longer need to be treated as orthogonal: the identity loss makes generated images more retrievable while improving realism, so the two tasks reinforce each other.
  • At 25.9M parameters and 39.2 FPS, the model is smaller and faster than the compared baselines while achieving the best FID, which makes retrieval-guided GANs plausible for real-time deployment.
  • The introduction of VIGOR-GEN (103,516 roughly centered urban image pairs) gives the community a harder benchmark with complex facades, narrow views, and occlusion.
  • The paper's claim that retrieval embeddings carry enough semantic signal implies that other cross-domain tasks with a pretrained metric embedding could be approached this way.

Reading between the lines

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

  • Editorial inference: the R@1 metric is computed with SAIG-D, a sibling model of the SAIG-S embedder used in training, so part of the retrieval gain could come from matching the quirks of one embedding family rather than genuine generalization; re-evaluating generated images with an independently trained geo-localization model would separate the two.
  • Editorial inference: the paper attributes the method's success to smoothness of the embedding space and contrasts it with the non-smooth instance-loss space of LPN; a sharper test would keep retrieval accuracy matched between two embedders and vary only space smoothness, then measure whether identity-loss training still helps.
  • Editorial inference: because the pipeline needs only a frozen embedder, swapping in a vision-language or text-aligned embedder would extend the same guidance mechanism to text-conditioned or instruction-based cross-view synthesis, which the paper does not explore.
  • Editorial inference: VIGOR-GEN is derived from VIGOR by selecting roughly centered pairs; using the non-centered pairs of VIGOR as a test set would probe whether the retrieval-guided identity loss is robust to the misalignment that occurs in real urban geo-localization, where the current method may need a differentiable alignment step.
Share X Bluesky LinkedIn Reddit HN

Signed reviews

No signed human review yet.

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

4 major / 6 minor

Summary. The paper proposes a retrieval-guided GAN framework for cross-view image synthesis. A fixed pre-trained cross-view retrieval network (SAIG-S) provides an embedding that is fused into a two-stage generator through an attentional AdaIN mechanism, along with style modulation, and an identity loss encourages the generated image to share view-invariant semantics with the target. The authors also introduce VIGOR-GEN, a new urban cross-view dataset derived from VIGOR. Experiments on CVUSA, CVACT, and VIGOR-GEN report state-of-the-art results in SSIM, PSNR, LPIPS, FID, and R@1, including large FID improvements over prior methods.

Significance. If the results are reliable, the paper makes a useful contribution: it shows that a retrieval embedding alone can provide a conditioning signal for cross-view synthesis without segmentation maps or polar-transformation preprocessing, and it introduces a more challenging urban dataset. The method is relatively lightweight and the paper includes ablations. However, two load-bearing problems currently prevent accepting the headline claims: the R@1 metric is evaluated with a sibling model of the embedder used in the training objective, making the retrieval-accuracy comparison with baselines unfair, and the main ablation table is inconsistent with the direction stated in the text. The lack of error bars further weakens the quantitative comparisons.

major comments (4)
  1. [Sec. 5.1 (Metrics), Eq. (9), Sec. 3.2] The reported R@1 metric is not an independent measure of retrieval quality. The identity loss in Eq. (9) uses R = SAIG-S (Sec. 3.2), and R@1 is evaluated with SAIG-D, another model from the same prior paper [46]. Since the generator is trained to maximize cosine similarity in SAIG-S space, and SAIG-S and SAIG-D are siblings trained with the same contrastive objective on the same data, the method is directly optimizing the criterion it is later scored on, whereas none of the baselines do so. The large R@1 gaps (e.g., 96.25 vs. 85.04 on CVUSA a2g) are therefore not evidence of generalizable cross-view correspondence. The LPN experiment in Table 5, which changes the embedder family, drops R@1 to 30.45, indicating that the R@1 number is strongly tied to the SAIG embedding family. To support the claim, the authors should evaluate with an independent retrieval model not used in training, or at least with an embedder from a different family, and should ablate the identity loss while keeping the evaluation embedder fixed.
  2. [Sec. 5.3, Table 4] The ablation study is internally inconsistent. The text states "We report variant models at the g2a direction on CVUSA," but the "Ours" row of Table 4 (SSIM 0.3702, PSNR 14.33, LPIPS 0.4302, FID 13.57, R@1 96.25) matches the a2g row of Table 2 (0.3706/14.33/0.4302/13.57/96.25), not the g2a row (0.2461/15.77/0.5181/41.65/95.14). If the table is at a2g, the text must say so; if the text is correct, the table is wrong. As written, the conclusions about the contribution of each component are not supported because the baseline for the claimed direction is missing.
  3. [Sec. 5.2, Tables 2–7] No error bars or significance tests are reported anywhere; all tables show single-run values. Given the stochasticity of GAN training and the small differences in some metrics (e.g., PSNR 14.33 vs. 13.91 on CVUSA a2g), the claimed improvements may not be statistically significant. The authors should report mean and standard deviation over at least three seeds for the main comparisons, or perform significance testing, before claiming state-of-the-art performance.
  4. [Sec. 4] VIGOR-GEN is a claimed contribution, but its construction is not described in sufficient detail. The text says it is derived from VIGOR and made roughly centered, but it does not specify the alignment/cropping procedure, how the 103,516 pairs were obtained, how the train/test split (51,366/51,250) was formed, or whether any filtering was applied. Without this information, the dataset cannot be reproduced and potential alignment errors or train/test leakage cannot be assessed.
minor comments (6)
  1. [Introduction] The sentence "We demonstrate how advances in retrieval techniques can beyond traditional retrieval scenarios" is missing a verb and should be revised.
  2. [Sec. 3.1] The text says "leverages a pre-trained and fixed retrieval model to identify view-invariant semantics and view-invariant semantics" — the second occurrence should presumably be "view-specific semantics."
  3. [Eq. (1)] The variance formula is miswritten: the term should be (X - mu_nc)^2, not (X - mu_nc^2). Please correct the notation.
  4. [Sec. 5.2] The claim "surpassing S2SP by 6 points in SSIM" is numerically incorrect; Table 2 shows SSIM 0.3706 vs. 0.3437, an improvement of 0.0269, not 6 points.
  5. [References] References [37] and [38] refer to the same paper (Zhai et al., CVPR 2017) and should be merged to avoid duplication.
  6. [Sec. 5.4, Figure 6] The claim that the SAIG embedding space is "smooth" is supported only by a convergence curve; this is not a formal notion of smoothness. Please clarify what is meant by smoothness and how it is measured.

Circularity Check

2 steps flagged · score 6.0 of 10

Headline R@1 gains are the identity-loss objective: Eq. 9 trains the generator against SAIG-S cosine similarity, and R@1 is then scored with SAIG-D, a sibling model from the same authors' prior paper; the retrieval metric is not an independent test.

  1. fitted input called prediction [Sec. 3.4, Eq. (9); Sec. 5.1, Metrics]
    "To ensure that the synthesized image has the same view-invariant semantics information as the target image, we use identity loss, which is defined as: L_id =1− cos(R(X_r),R(X_t)) + 1− cos(R(X'_r),R(X'_t)) (9) ... R denotes the pre-trained retrieval network as in Sec. 3.2. ... we reported the Recall@1 (R@1) in our experiment using another cross-view image retrieval model SAIG-D [46]."

    Eq. 9 is exactly a retrieval loss: it drives the generated image X_r to maximize cosine similarity with the target X_t in the SAIG-S embedding space. The headline R@1 metric then measures retrieval rank in the same kind of embedding similarity, using SAIG-D, another model from the same prior paper [46] that provides SAIG-S. The reported retrieval accuracy is therefore the training objective's own score transferred to a sibling checkpoint, not an independent test of cross-view correspondence. The paper interprets R@1 as evidence that generated images 'describe the same location,' but the generator was explicitly trained to make SAIG-S embeddings agree; SAIG-D is a same-family check, and no experiment evaluates R@1 with an independent retrieval embedder while keeping the identity loss fixed.

  2. self citation load bearing [Sec. 3.2; Sec. 5.4]
    "We utilize its shallower version SAIG-S [46] here. This retrieval network can settle visual differences and directly embed images from different views into a smooth space. ... By using retrieval loss, the embedding space remains smooth, aiding in better cross-view retrieval."

    The central premise that the fixed embedder provides a smooth, semantically meaningful space—and hence that Eq. 9 gives reliable gradients for 'view-invariant semantics'—is imported from the authors' own prior work [46] rather than independently verified in this paper. That asserted smoothness is then used to explain the identity loss's behavior and to interpret R@1 as semantic correspondence. The only alternative embedder tested (LPN) is from a different group, but the final R@1 evaluation still uses the SAIG family, so the load-bearing smoothness assumption is never checked with an independent retrieval model.

full rationale

The synthesis-quality claims (FID, SSIM, PSNR, LPIPS) use standard external feature extractors and are not circular; those comparisons retain independent content. The circularity is concentrated in the R@1 claim: the identity loss in Eq. 9 optimizes the same cosine-similarity retrieval objective that R@1 reports, using SAIG-S and SAIG-D from the authors' own prior paper [46]. Because R@1 is a headline result, highlighted in the abstract as 'particularly in retrieval accuracy (R@1),' the evaluation partially reduces to the training objective. A fully independent retrieval embedder, or a cross-family evaluation with a non-SAIG model, would resolve the concern. Score 6 reflects partial circularity rather than total collapse of the paper's contribution.

Assumptions & free parameters 4 free parameters · 5 assumptions · 0 invented entities

The method rests on the smoothness of a fixed retrieval embedding and on the validity of the derived VIGOR-GEN dataset and of R@1 evaluation with a model from the same paper family. Loss weights are hand-set.

free parameters (4)
  • lambda_id = 10
    Identity loss weight in Eq. 12; hand-set. It controls how strongly the generator matches the retrieval embedding of the target, which directly affects the R@1 metric.
  • lambda_rec = 50
    L1 reconstruction weight in Eq. 12; hand-set.
  • lambda_perc = 50
    LPIPS perceptual loss weight in Eq. 12; hand-set.
  • lambda_div = 0.1 (CVUSA/CVACT), 1 (VIGOR-GEN)
    Diversity loss weight in Eq. 12; hand-set differently across datasets.
assumptions (5)
  • domain assumption The pre-trained retrieval embedder SAIG-S (adopted from [46]) maps cross-view images into a smooth embedding space where cosine distance reflects view-invariant semantic similarity.
    Sec. 3.2 and Sec. 5.4 argue smoothness is key; the comparison with LPN is suggestive but not a proof of the exact property needed for the identity loss.
  • domain assumption The identity loss gradient is a reliable, non-degenerate learning signal that improves genuine cross-view correspondence rather than exploiting artifacts of the embedder.
    Eq. 9 and Sec. 5.4; if the loss can be minimized by outputting images that fool the embedder, the synthesis gains and R@1 would not reflect true correspondence.
  • domain assumption The evaluation retrieval model SAIG-D from the same reference [46] yields a fair, independent measure of whether generated images depict the same location as the target.
    Sec. 5.1 defines R@1 using SAIG-D; however, SAIG-D is from the same paper family as the embedder used in the identity loss, so this independence is questionable.
  • ad hoc to paper The VIGOR-GEN dataset is a valid centrally aligned cross-view dataset derived from VIGOR; the alignment/cropping procedure is assumed to produce correct ground-truth pairs.
    Sec. 4 states it is 'derived' and 'mainly extended on VIGOR' but gives no concrete alignment method; the main VIGOR-GEN results depend on this.
  • standard math Adversarial training with spectral normalization and standard losses converges to a usable local optimum for image synthesis.
    Standard GAN practice, not the focus of this paper.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Retrieval-guided Cross-view Image Synthesis." pith.science (2026). https://pith.science/paper/YWHRYUCW

@misc{pith2026241119510,
  author       = {Pith},
  title        = {Pith review of: Retrieval-guided Cross-view Image Synthesis},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/YWHRYUCW}},
  note         = {Machine review of arXiv:2411.19510}
}
read the original abstract

Information retrieval techniques have demonstrated exceptional capabilities in identifying semantic similarities across diverse domains through robust feature representations. However, their potential in guiding synthesis tasks, particularly cross-view image synthesis, remains underexplored. Cross-view image synthesis presents significant challenges in establishing reliable correspondences between drastically different viewpoints. To address this, we propose a novel retrieval-guided framework that reimagines how retrieval techniques can facilitate effective cross-view image synthesis. Unlike existing methods that rely on auxiliary information, such as semantic segmentation maps or preprocessing modules, our retrieval-guided framework captures semantic similarities across different viewpoints, trained through contrastive learning to create a smooth embedding space. Furthermore, a novel fusion mechanism leverages these embeddings to guide image synthesis while learning and encoding both view-invariant and view-specific features. To further advance this area, we introduce VIGOR-GEN, a new urban-focused dataset with complex viewpoint variations in real-world scenarios. Extensive experiments demonstrate that our retrieval-guided approach significantly outperforms existing methods on the CVUSA, CVACT and VIGOR-GEN datasets, particularly in retrieval accuracy (R@1) and synthesis quality (FID). Our work bridges information retrieval and synthesis tasks, offering insights into how retrieval techniques can address complex cross-domain synthesis challenges.

Figures

Figures reproduced from arXiv: 2411.19510 by the authors.

Figure 1
Figure 1. Illustrating view-invariant and view-specific seman￾tics in cross-view image synthesis between aerial and ground views. Cross-view image synthesis aims to generate an image from a novel viewpoint, given one input image from different viewpoint, such as transforming an aerial (bird’s-eye) view into a ground (street) view. [8, 23, 28, 30, 32, 35]. This technique can benefit a wide range of applications, from autonomou… view at source ↗
Figure 2
Figure 2. Illustration of our network architecture. left: our network consists of a structure generator, a facade generator, a [PITH_FULL_IMAGE:figures/full_fig_p004_2.png] view at source ↗
Figure 3
Figure 3. Comparison of current methods at a2g direction on CVACT. [PITH_FULL_IMAGE:figures/full_fig_p007_3.png] view at source ↗
Figures from the paper (4 more)
Figure 4
Figure 4. Figure 4: Comparison of g2a (ground-to-aerial) synthesis on [PITH_FULL_IMAGE:figures/full_fig_p007_4.png]
Figure 5
Figure 5. Figure 5: Visualization of the weight map 𝑀 on VIGOR-GEN [PITH_FULL_IMAGE:figures/full_fig_p008_5.png]
Figure 6
Figure 6. Figure 6: The curve of ID loss in generator using different [PITH_FULL_IMAGE:figures/full_fig_p008_6.png]
Figure 7
Figure 7. Figure 7: Comparison of images generated by models using different Residual Blocks. [PITH_FULL_IMAGE:figures/full_fig_p009_7.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

49 extracted references · 28 canonical work pages

  1. [46]

    Yingying Zhu, Hongji Yang, Yuxin Lu, and Qiang Huang. 2023. Simple, Effective and General: A New Backbone for Cross-view Image Geo-localization. arXiv e-prints, Article arXiv:2302.01572 (Feb. 2023), arXiv:2302.01572 pages. doi:10. 48550/arXiv.2302.01572 arXiv:2302.01572 [cs.CV]

    arXiv doi:10.48550/arxiv.2302.01572 2023

  2. [1]

    Benjamin Bischke, Damian Borth, Christian Schulze, and Andreas Dengel. 2016. Contextual enrichment of remote-sensed events with social media streams. In Proceedings of the 24th ACM international conference on Multimedia . 1077–1081

    work page 2016

  3. [2]

    Tim Brooks, Aleksander Holynski, and Alexei A Efros. 2023. Instructpix2pix: Learning to follow image editing instructions. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition . 18392–18402

    2023

  4. [3]

    Yunjey Choi, Youngjung Uh, Jaejun Yoo, and Jung-Woo Ha. 2020. Stargan v2: Diverse image synthesis for multiple domains. In Proceedings of the IEEE/CVF conference on computer vision and pattern recognition . 8188–8197

    work page 2020

  5. [4]

    Florinel-Alin Croitoru, Vlad Hondru, Radu Tudor Ionescu, and Mubarak Shah

  6. [5]

    Harm De Vries, Florian Strub, Jérémie Mary, Hugo Larochelle, Olivier Pietquin, and Aaron C Courville. 2017. Modulating early visual processing by language. Advances in Neural Information Processing Systems 30 (2017)

    work page 2017

  7. [6]

    Martin Heusel, Hubert Ramsauer, Thomas Unterthiner, Bernhard Nessler, and Sepp Hochreiter. 2017. Gans trained by a two time-scale update rule converge to a local nash equilibrium. Advances in neural information processing systems 30 (2017)

    2017

  8. [7]

    Xun Huang and Serge Belongie. 2017. Arbitrary style transfer in real-time with adaptive instance normalization. In Proceedings of the IEEE international conference on computer vision . 1501–1510

    2017

Show all 49 references
  1. [8]

    Phillip Isola, Jun-Yan Zhu, Tinghui Zhou, and Alexei A Efros. 2017. Image-to- image translation with conditional adversarial networks. In Proceedings of the IEEE conference on computer vision and pattern recognition . 1125–1134

    2017

  2. [9]

    Tero Karras, Miika Aittala, Janne Hellsten, Samuli Laine, Jaakko Lehtinen, and Timo Aila. 2020. Training generative adversarial networks with limited data. Advances in neural information processing systems 33 (2020), 12104–12114

    2020

  3. [10]

    Tero Karras, Samuli Laine, and Timo Aila. 2019. A style-based generator ar- chitecture for generative adversarial networks. In Proceedings of the IEEE/CVF conference on computer vision and pattern recognition . 4401–4410

    2019

  4. [11]

    Tero Karras, Samuli Laine, Miika Aittala, Janne Hellsten, Jaakko Lehtinen, and Timo Aila. 2020. Analyzing and improving the image quality of stylegan. In Proceedings of the IEEE/CVF conference on computer vision and pattern recognition . 8110–8119

    2020

  5. [12]

    Diederik P Kingma and Jimmy Ba. 2014. Adam: A method for stochastic opti- mization. arXiv preprint arXiv:1412.6980 (2014)

    2014 arXiv

  6. [13]

    Hsin-Ying Lee, Hung-Yu Tseng, Qi Mao, Jia-Bin Huang, Yu-Ding Lu, Maneesh Singh, and Ming-Hsuan Yang. 2020. Drit++: Diverse image-to-image translation via disentangled representations. International Journal of Computer Vision 128 (2020), 2402–2417

    2020

  7. [14]

    Liu Liu and Hongdong Li. 2019. Lending orientation to neural networks for cross- view geo-localization. In Proceedings of the IEEE/CVF conference on computer vision and pattern recognition . 5624–5633

    2019

  8. [15]

    Xiaohu Lu, Zuoyue Li, Zhaopeng Cui, Martin R Oswald, Marc Pollefeys, and Rongjun Qin. 2020. Geometry-aware satellite-to-ground image synthesis for urban areas. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition. 859–867

    2020

  9. [16]

    Jisan Mahmud, True Price, Akash Bapat, and Jan-Michael Frahm. 2020. Boundary- aware 3D building reconstruction from a single overhead image. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition . 441–451

    2020

  10. [17]

    Qi Mao, Hsin-Ying Lee, Hung-Yu Tseng, Siwei Ma, and Ming-Hsuan Yang. 2019. Mode seeking generative adversarial networks for diverse image synthesis. In Proceedings of the IEEE/CVF conference on computer vision and pattern recognition . 1429–1437

    2019

  11. [18]

    Gellért Máttyus, Wenjie Luo, and Raquel Urtasun. 2017. Deeproadmapper: Ex- tracting road topology from aerial images. InProceedings of the IEEE international conference on computer vision . 3438–3446

    2017

  12. [19]

    Takeru Miyato, Toshiki Kataoka, Masanori Koyama, and Yuichi Yoshida. 2018. Spectral normalization for generative adversarial networks. arXiv preprint arXiv:1802.05957 (2018)

    2018 arXiv

  13. [20]

    Taesung Park, Ming-Yu Liu, Ting-Chun Wang, and Jun-Yan Zhu. 2019. Semantic image synthesis with spatially-adaptive normalization. In Proceedings of the IEEE/CVF conference on computer vision and pattern recognition . 2337–2346

    2019

  14. [21]

    Ming Qian, Jincheng Xiong, Gui-Song Xia, and Nan Xue. 2023. Sat2density: Faithful density learning from satellite-ground image pairs. In Proceedings of the IEEE/CVF International Conference on Computer Vision . 3683–3692

    2023

  15. [22]

    Aditya Ramesh, Prafulla Dhariwal, Alex Nichol, Casey Chu, and Mark Chen

  16. [23]

    Krishna Regmi and Ali Borji. 2018. Cross-view image synthesis using condi- tional gans. In Proceedings of the IEEE conference on Computer Vision and Pattern Recognition. 3501–3510

    2018

  17. [24]

    Elad Richardson, Yuval Alaluf, Or Patashnik, Yotam Nitzan, Yaniv Azar, Stav Shapiro, and Daniel Cohen-Or. 2021. Encoding in style: a stylegan encoder for image-to-image translation. InProceedings of the IEEE/CVF conference on computer vision and pattern recognition . 2287–2296

    2021

  18. [25]

    Robin Rombach, Andreas Blattmann, Dominik Lorenz, Patrick Esser, and Björn Ommer. 2022. High-resolution image synthesis with latent diffusion models. In Proceedings of the IEEE/CVF conference on computer vision and pattern recognition . 10684–10695

    2022

  19. [26]

    Chitwan Saharia, William Chan, Saurabh Saxena, Lala Li, Jay Whang, Emily L Denton, Kamyar Ghasemipour, Raphael Gontijo Lopes, Burcu Karagol Ayan, Tim Salimans, et al. 2022. Photorealistic text-to-image diffusion models with deep language understanding. Advances in Neural Infor...

    2022

  20. [27]

    Axel Sauer, Tero Karras, Samuli Laine, Andreas Geiger, and Timo Aila. 2023. Stylegan-t: Unlocking the power of gans for fast large-scale text-to-image syn- thesis. arXiv preprint arXiv:2301.09515 (2023)

    2023 arXiv

  21. [28]

    Yujiao Shi, Dylan Campbell, Xin Yu, and Hongdong Li. 2022. Geometry-Guided Street-View Panorama Synthesis From Satellite Imagery. IEEE Transactions on Pattern Analysis and Machine Intelligence 44, 12 (2022), 10009–10022. doi:10.1109/ TPAMI.2022.3140750

    2022

  22. [29]

    Yujiao Shi, Liu Liu, Xin Yu, and Hongdong Li. 2019. Spatial-aware feature aggrega- tion for image based cross-view geo-localization. Advances in Neural Information Processing Systems 32 (2019)

    2019

  23. [30]

    Hao Tang, Dan Xu, Nicu Sebe, Yanzhi Wang, Jason J Corso, and Yan Yan. 2019. Multi-channel attention selection gan with cascaded semantic guidance for cross- view image translation. In Proceedings of the IEEE/CVF conference on computer vision and pattern recognition . 2417–2426

    2019

  24. [31]

    Ming Tao, Hao Tang, Fei Wu, Xiao-Yuan Jing, Bing-Kun Bao, and Changsheng Xu. 2022. Df-gan: A simple and effective baseline for text-to-image synthesis. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition. 16515–16525

    2022

  25. [32]

    Aysim Toker, Qunjie Zhou, Maxim Maximov, and Laura Leal-Taixé. 2021. Coming down to earth: Satellite-to-street view synthesis for geo-localization. In Proceed- ings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition . 6488–6497

    2021

  26. [33]

    Hugo Touvron, Matthieu Cord, Alexandre Sablayrolles, Gabriel Synnaeve, and Hervé Jégou. 2021. Going deeper with image transformers. In Proceedings of the IEEE/CVF international conference on computer vision . 32–42

    2021

  27. [34]

    Tingyu Wang, Zhedong Zheng, Chenggang Yan, Jiyong Zhang, Yaoqi Sun, Bolun Zhenga, and Yi Yang. 2021. Each part matters: Local patterns facilitate cross-view geo-localization. TCSVT (2021)

    2021

  28. [35]

    Songsong Wu, Hao Tang, Xiao-Yuan Jing, Haifeng Zhao, Jianjun Qian, Nicu Sebe, and Yan Yan. 2022. Cross-view panorama image synthesis. IEEE Transactions on Multimedia (2022)

    2022

  29. [36]

    Shuai Yang, Liming Jiang, Ziwei Liu, and Chen Change Loy. 2022. Pastiche master: Exemplar-based high-resolution portrait style transfer. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition . 7693–7702

    2022

  30. [37]

    Menghua Zhai, Zachary Bessinger, Scott Workman, and Nathan Jacobs. 2017. Predicting ground-level scene layout from aerial imagery. In Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition . 867–875

    2017

  31. [38]

    Menghua Zhai, Zachary Bessinger, Scott Workman, and Nathan Jacobs. 2017. Predicting ground-level scene layout from aerial imagery. In CVPR. 867–875

    2017

  32. [39]

    Lvmin Zhang, Anyi Rao, and Maneesh Agrawala. 2023. Adding conditional con- trol to text-to-image diffusion models. InProceedings of the IEEE/CVF International Conference on Computer Vision . 3836–3847

    2023

  33. [40]

    Richard Zhang, Phillip Isola, Alexei A Efros, Eli Shechtman, and Oliver Wang

  34. [41]

    Shengyu Zhao, Zhijian Liu, Ji Lin, Jun-Yan Zhu, and Song Han. 2020. Differentiable augmentation for data-efficient gan training. Advances in Neural Information Processing Systems 33 (2020), 7559–7570

    2020

  35. [42]

    Zhedong Zheng, Yunchao Wei, and Yi Yang. 2020. University-1652: A multi-view multi-source benchmark for drone-based geo-localization. In Proceedings of the 28th ACM international conference on Multimedia . 1395–1403

    2020

  36. [43]

    Peihao Zhu, Rameen Abdal, Yipeng Qin, and Peter Wonka. 2020. Sean: Image synthesis with semantic region-adaptive normalization. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition . 5104–5113

    2020

  37. [44]

    Sijie Zhu, Taojiannan Yang, and Chen Chen. 2021. Vigor: Cross-view image geo-localization beyond one-to-one retrieval. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition . 3640–3649

    2021

  38. [45]

    Yingying Zhu, Shihai Chen, Xiufan Lu, and Jianyong Chen. 2023. Cross-view Image Synthesis from a Single Image with Progressive Parallel GAN. IEEE Transactions on Geoscience and Remote Sensing (2023). Retrieval-guided Cross-view Image Synthesis Conference’17, July 2017, Washing...

    2023

  39. [2018]

    In Proceedings of the IEEE conference on computer vision and pattern recognition

    The unreasonable effectiveness of deep features as a perceptual metric. In Proceedings of the IEEE conference on computer vision and pattern recognition . 586–595

  40. [2022]

    arXiv preprint arXiv:2204.06125 1, 2 (2022), 3

    Hierarchical text-conditional image generation with clip latents. arXiv preprint arXiv:2204.06125 1, 2 (2022), 3

    2022 arXiv

  41. [2023]

    IEEE Transactions on Pattern Analysis and Machine Intelligence (2023)

    Diffusion models in vision: A survey. IEEE Transactions on Pattern Analysis and Machine Intelligence (2023)

    2023

Pith tools

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