Pith. sign in

REVIEW 4 major objections 5 minor 58 references

GeoCore-9B: Towards Geo-Aware Generative Foundation Models in Earth Observation

T0 review · 4 major / 5 minor · reviewed 2026-08-15 · deepseek-v4-flash

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

Pith's one-line read GeoCore-9B claims a 9-billion-parameter diffusion transformer trained from scratch on satellite data, with text and geospatial conditioning, beats fine-tuned natural-image priors on fidelity and geographic accuracy.

desk verdict A serious, unusually transparent EO-native 9B generative foundation model with real supporting evidence for its GSA mechanism, but the downstream SOTA claims currently rest on a missing leakage check. read the letter →

arxiv 2608.01896 v1 pith:S3WFTMHU submitted 2026-08-03 cs.CV

Jeonghyeok Do , Munchurl Kim This is my paper · ORCID
classification cs.CV
keywords earthobservationgenerativefoundationmodeldiffusiontransformerflowmatchinggeospatialconditioninggroundsampledistancecloudremovalSAR-to-opticaltranslation
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 aims to establish that a generative foundation model for Earth observation can be built from scratch, without natural-image priors, and that at sufficient scale it becomes both more geo-aware and more practically transferable than existing fine-tuned EO generators. It reports GeoCore-9B, a 9-billion-parameter Flow Matching-based Diffusion Transformer pre-trained on a 10-million-pair global EO dataset with text, ground-sample-distance, latitude, and longitude conditioning, as the first model of its scale to do so. The key claimed mechanism is a Geospatial Semantic Alignment loss: during training, a frozen satellite-specialist teacher's dense structural features are distilled into intermediate DiT representations, improving structural fidelity and convergence without adding any inference cost. If the claims hold, a single large EO-trained generative prior can handle text-driven generation, location-only generation, cloud removal, and SAR-to-optical translation better than specialized systems built for each task.

What carries the argument

The central object is a Flow Matching-based Diffusion Transformer (DiT), a transformer that learns to predict the straight-line velocity $v = z_1 - z_0$ along the linear latent trajectory $z_t = (1-t)z_0 + t z_1$ connecting Gaussian noise to a frozen VAE-encoded image, with conditioning injected through adaptive layer normalization (AdaLN). The conditions are a global text embedding, token-level text embeddings, and sinusoidally encoded geospatial scalars (ground sample distance, latitude, longitude) summed through MLPs; 3D rotary positional embeddings keep text and image tokens jointly position-aware. The paper's named new component is the Geospatial Semantic Alignment (GSA) loss: at the 8th DiT block, projected intermediate features are pushed toward dense features from a frozen satellite-specialist teacher with an $\ell^2$ loss at weight $\mu = 0.5$. Because the teacher and projection are used only in training, GSA shapes the latent trajectory without adding inference time or parameters.

What would settle it

The decisive control is a same-scale natural-image-pretrained diffusion transformer fine-tuned on the same 10M-pair EO corpus with the same LoRA recipe; if it matches or beats GeoCore-9B's FID, PSNR, and HF-SCC numbers, then the claim that from-scratch EO training and geospatial conditioning drive the improvement fails. A second check is to run the frozen VAE's reconstruction on small sparse targets and extreme high-frequency bands; the paper's own Appendix B reports strong averages but concedes it does not guarantee preservation of these, so a failure there would clamp the model's achievable detail below what the reported metrics suggest.

Watch

Extended reading notes

Core claim

GeoCore-9B is presented as the first 9-billion-parameter generative foundation model trained from scratch exclusively on Earth observation data, and the paper's central claim is that this from-scratch training with native geospatial conditioning, rather than inherited natural-image priors, is what yields state-of-the-art visual fidelity and geographic structural accuracy. The model conditions on text, ground sample distance, latitude, and longitude, and the authors report that it can generate location-appropriate terrain from coordinates alone, without any text prompt. They further report that after freezing the backbone and tuning only low-rank adapters (LoRA), GeoCore-9B outperforms task-specific specialists on cloud removal in PSNR and SSIM, and achieves the best FID and highest high-frequency spatial correlation on SAR-to-optical translation, remaining competitive in SSIM. The supporting design claim is that the Geospatial Semantic Alignment loss, which distills structural Earth-surface priors from a frozen specialist teacher during training, stabilizes from-scratch convergence and explains a measured share of the gains in the paper's matched ablations.

Load-bearing premise

The whole result leans on the frozen natural-image VAE preserving enough satellite-specific detail: the paper's Appendix B audit reports 31.48 dB PSNR and 0.710 high-frequency correlation on a 100,000-image sample, but concedes it does not guarantee preservation of every tiny target or extreme frequency band, so any detail the encoder discards is permanently unavailable to both generation and downstream tasks.

Editorial extensions

If this is right

  • An EO generative prior trained from scratch can shed the perspective-projection and center-object biases that natural-image fine-tuning carries, making orthographic structure and scale-awareness inherent rather than patched on.
  • Because geospatial metadata is natively conditioned, the model can generate useful location-specific imagery from coordinates alone; this is the text-free capability the paper demonstrates.
  • A frozen EO-trained backbone plus LoRA adapters is enough to reach or beat task-specific specialists on cloud removal and SAR-to-optical translation; the paper's reported PSNR/SSIM and FID/HF-SCC results support that.
  • The GSA loss improves convergence and structural fidelity without inference cost: in the paper's matched 9B ablation, removing it degrades FID, PSNR, SSIM, and boundary sharpness.
  • The frozen natural-image VAE's reconstruction quality sets an upper bound on achievable detail, so replacing it with an EO-specific latent encoder, the paper's stated future direction, is the next lever for further gains.

Reading between the lines

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

  • If the from-scratch advantage is real, the next scaling law for EO generation should show continued gains from parameter count and data alone; a plateau caused by the frozen latent encoder would be observable before any architecture change.
  • The coordinate-only generations imply the weights encode a coarse global geography; a region-by-region retrieval audit would distinguish genuine geospatial generalization from localized memorization, since the paper's near-duplicate test uses only global features at one threshold.
  • The SAR-to-optical result suggests an optical structural prior can help an input modality the GSA teacher never saw, so a cheap extension is to apply the same alignment loss to other non-optical inputs and see whether the transfer persists.
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 / 5 minor

Summary. GeoCore-9B proposes a 9-billion-parameter Flow Matching-based diffusion transformer trained from scratch on the Git-10M satellite RGB dataset, with conditioning on text, ground sample distance, latitude, and longitude, plus a Geospatial Semantic Alignment (GSA) loss that distills structural priors from a frozen DINOv3-Sat teacher during training. The paper evaluates the model on RSICD text-to-image generation, cloud removal on Sen2-MTC, SAR-to-optical translation on QXS-SAROPT, and includes a matched 9B GSA ablation, a frozen-VAE audit, metadata intervention tests, near-duplicate retrieval, and frozen-feature linear probes.

Significance. If the claims hold, the paper makes a strong contribution: it is among the first to train a 9B-scale generative model from scratch on EO data rather than fine-tuning natural-image priors, and the GSA loss adds no inference cost. The matched 9B ablation (Table 4), the explicit VAE audit (Appendix B), and the paired metadata interventions (Table 5) are exemplary in directly testing mechanistic claims, and the paper's candid limitation statements are a significant strength. The frozen-feature probes (Appendix D) provide additional evidence of transferable representations. However, the headline structural-accuracy metric is internally inconsistent between tables, the downstream claims are not yet protected against pretraining leakage, and the absence of error bars leaves the margins of key comparisons unquantified. These issues currently prevent acceptance.

major comments (4)
  1. [§4.6, Table 2(b) vs §C.1, Table 4] The HF-SCC values are inconsistent across tables: Table 2(b) reports GeoCore-9B HF-SCC of 0.3360 on QXS-SAROPT, while the matched 9B ablation in Table 4 reports 0.0163 for the same w/ GSA condition, with the note that Table 4 uses a 'corrected, baseline-consistent definition.' This implies the main text's headline structural-accuracy value rests on a different, apparently uncorrected protocol. Because the claimed geographic structural SOTA and the very large margin over C-DiffSET (0.336 vs 0.0108) depend on this metric, the discrepancy is load-bearing. Please recompute all HF-SCC values under a single definition, report them consistently in both tables, and justify the chosen definition.
  2. [§C.3 and §4.6] The near-duplicate retrieval protocol checks only 500 coordinate-only generations against the Git-10M pretraining corpus, not the Sen2-MTC cloud-free references or the QXS-SAROPT optical references. Because Git-10M is a global corpus and these downstream datasets have known sensor footprints, overlap is plausible; if evaluation reference images (or transformed variants) appear in pretraining, the LoRA-adapted model could exploit memorized location-specific appearance, inflating PSNR, SSIM, and HF-SCC and undermining the 'transferable generative prior' interpretation. The appendix itself concedes that the protocol 'cannot exclude localized, transformed, or other forms of memorization.' Please run the same near-duplicate retrieval between the downstream evaluation sets and Git-10M at the stated threshold, and report the overlap. If overlap is found, the downstream analysis must be conditioned on non-overlapping splits.
  3. [Tables 1, 2, and 4] All main results are single-run point estimates without error bars or significance tests. Several key comparisons are close, such as LPIPS 0.256 vs 0.255 for EMRDM in Table 2(a) and SSIM 0.370 vs 0.372 for C-DiffSET in Table 2(b), so the claimed margins fall within plausible run-to-run noise. This is especially important for the central GSA mechanism claim in Table 4, where the reported improvements (e.g., FID 18.82 vs 28.43 on RSICD) have no associated variance. Please provide at least three seeds with means and standard deviations (or confidence intervals) for the main tables, and apply a statistical test to the matched GSA ablation.
  4. [§1 and §4.6] The downstream comparisons in Tables 2(a) and 2(b) are not controlled: task-specific baselines use their native architectures, losses, input modalities (multispectral, multi-temporal, full fine-tuning), while GeoCore-9B is adapted via LoRA on a frozen RGB-only backbone with a generic Flow Matching objective. The contribution statement in Section 1 claims GeoCore-9B 'outperforms or remains competitive compared to task-specific specialist methods' (and the abstract states it establishes new SOTA on cloud removal and SAR-to-optical), but Section G acknowledges that the experiments do not decompose the margin to external baselines. Please either re-run a matched set of baselines under the same LoRA/RGB protocol, or temper the abstract and contribution claims to 'competitive with' and explicitly list protocol differences in the main text.
minor comments (5)
  1. [Abstract and §G] The abstract says the model is 'trained from scratch exclusively on EO data,' but Section G correctly notes that the system uses a pretrained natural-image VAE and pretrained text encoders. Please rephrase to specify that the DiT backbone is trained from scratch on EO data, while the surrounding system retains natural-image components.
  2. [Table 1] The Text2Earth row has no Inception Score; if the value is not reported in the original source, state 'not reported' in the table caption or a footnote rather than leaving the cell empty.
  3. [§4.3] The 'text-free' coordinate-only comparison in Fig. 4 includes CRS-Diff but not Text2Earth; consider including Text2Earth in this qualitative comparison or justifying its omission, since Text2Earth is the more direct text-to-image comparator.
  4. [§4.4 and Fig. 6] The description says GSA 'consistently reduces FID on a 10K Git-10M subset,' but the figure only shows four training iterations; please specify the evaluation sample size, the y-axis scale, and whether the evaluation subset is fixed across iterations.
  5. [§3.3, Eq. (4)] The GSA loss applies at layer k=8 only; the paper does not explain how k was chosen. A sentence on the selection criterion (e.g., a small grid search or prior REPA-style practice) would improve reproducibility.

Circularity Check

0 steps flagged · score 0.0 of 10

No construction-level circularity: the Flow Matching objective, geospatial conditioning, and GSA ablation are self-contained; self-citations and appendix metric caveats do not reduce any claim to its inputs.

full rationale

GeoCore-9B's derivation chain is not circular. The core objective (Eq. 5) is a standard Flow Matching regression to the target velocity v = z1 - z0, with no fitted constant later renamed as a prediction; text, GSD, latitude, and longitude are direct conditioning inputs, not outputs of the evaluation metrics. The GSA loss (Eq. 4) distills a frozen DINOv3-Sat teacher during training, but the headline evaluations (IS, FID, CLIP, PSNR, SSIM, LPIPS, HF-SCC) are computed independently of that teacher, and Appendix C.1 isolates GSA with a matched 9B ablation, so the contribution is empirically demonstrated rather than assumed. External benchmarks (RSICD, Sen2-MTC, QXS-SAROPT) and many non-author baselines support the SOTA claims. The only self-cited baseline, C-DiffSET [2] in Table 2(b), is not load-bearing: it is one of several baselines, and GeoCore-9B is actually worse on LPIPS (0.377 vs 0.293), so the comparison does not reduce to a favorable self-citation. Caveats that should be weighed separately: Table 4's caption 'HF-SCC uses the corrected, baseline-consistent definition' indicates the main-text HF-SCC gap is sensitive to metric definition; Appendix B concedes the frozen VAE 'does not guarantee preservation of every tiny target or high-frequency structure'; and Appendix C.3 admits the near-duplicate protocol 'cannot exclude localized, transformed, or other forms of memorization.' These are correctness, leakage, and measurement-consistency risks for the transferability claim, not circular derivations from inputs to outputs.

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

The central claim rests on established generative architecture and several domain assumptions; no new physical entities are introduced. The main contributions are engineering scale and a novel alignment objective.

free parameters (5)
  • GSA loss weight mu = 0.5
    Controls strength of teacher alignment in Eq. (5); set empirically, no sensitivity analysis reported.
  • Classifier-free guidance scale w = 4.0
    Inference-time setting; no sweep reported.
  • LoRA rank and scale = r=64, alpha=128
    Downstream adaptation hyperparameters; no sweep reported.
  • Condition dropout probability p_cfg = 0.1
    Dropout for text and metadata conditions during training; no sensitivity study reported.
  • High-quality subset quality threshold = 4.8
    Used for post-pretraining refinement of Git-10M, following Text2Earth [23]; adopted without re-tuning.
assumptions (5)
  • standard math The Flow Matching formulation (Eq. 1) provides a valid and stable generative objective for latent image modeling.
    Established background from Lipman et al. [22] and Liu et al. [25]; the paper does not modify the objective.
  • domain assumption The frozen FLUX VAE preserves enough satellite-specific structure for the 256x256 tasks.
    Appendix B audits reconstruction at 31.48 dB PSNR and HF-SCC 0.710, but also concedes it cannot guarantee preservation of every tiny target or high-frequency structure.
  • domain assumption DINOv3-Sat features are a reliable teacher for Earth surface structural fidelity.
    Eq. (4) uses this teacher without a study of its failure modes; the matched ablation only shows an aggregate benefit.
  • domain assumption Git-10M metadata (GSD, latitude, longitude) is accurate enough to learn geo-aware conditioning.
    The model conditions on this metadata and coordinate-only results assume it is trustworthy; no independent metadata quality audit is reported.
  • domain assumption The evaluation metrics (IS, FID, CLIP, PSNR, SSIM, LPIPS, HF-SCC) capture the relevant aspects of EO generation quality.
    These are standard benchmarks from prior work; the paper provides no uncertainty quantification and no direct metric for geographic structural accuracy.

how reviews work

0 comments
Cite this review

Pith. "Pith review of GeoCore-9B: Towards Geo-Aware Generative Foundation Models in Earth Observation." pith.science (2026). https://pith.science/paper/S3WFTMHU

@misc{pith2026260801896,
  author       = {Pith},
  title        = {Pith review of: GeoCore-9B: Towards Geo-Aware Generative Foundation Models in Earth Observation},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/S3WFTMHU}},
  note         = {Machine review of arXiv:2608.01896}
}
read the original abstract

Existing generative models for earth observation (EO) predominantly rely on fine-tuning natural image priors, which limits their scalability and introduces perspective biases that conflict with geospatial constraints. To address this, we introduce GeoCore-9B, a 9-billion-parameter generative foundation model, which is the first of its scale to be trained from scratch exclusively on EO data. Unlike previous EO foundation models, GeoCore-9B is built upon a Flow Matching-based Diffusion Transformer (DiT) and natively conditions generation on text descriptions and continuous geospatial metadata, including ground sample distances, latitudes, and longitudes. To overcome the convergence and spatial disorientation challenges of training at this scale, we propose a Geospatial Semantic Alignment loss. This objective distills structural Earth surface priors (e.g., terrain and urban areas) from a frozen specialist teacher network, constraining the diffusion latent trajectory during training without adding inference overhead. Pre-trained on the global-scale Git-10M dataset, GeoCore-9B demonstrates strong downstream versatility. Beyond standard proxy generative tasks, we show that GeoCore-9B can be effectively adapted for practical EO applications, including highly challenging tasks such as cloud removal and SAR-to-optical cross-modal translation. Extensive evaluations confirm that GeoCore-9B establishes new state-of-the-art performance in both visual fidelity and geographic structural accuracy.

Figures

Figures reproduced from arXiv: 2608.01896 by the authors.

Figure 1
Figure 1. Overview of GeoCore-9B. GeoCore-9B trains a Flow Matching-based DiT in the latent [PITH_FULL_IMAGE:figures/full_fig_p003_1.png] view at source ↗
Figure 2
Figure 2. Qualitative comparison of text-conditioned generation. Given various text descriptions, [PITH_FULL_IMAGE:figures/full_fig_p005_2.png] view at source ↗
Figure 3
Figure 3. Qualitative comparison across varying ground sample distances (GSD). GeoCore-9B [PITH_FULL_IMAGE:figures/full_fig_p006_3.png] view at source ↗
Figures from the paper (11 more)
Figure 4
Figure 4. Figure 4: Qualitative comparison of text-free generation guided solely by latitude and longitude [PITH_FULL_IMAGE:figures/full_fig_p007_4.png]
Figure 5
Figure 5. Figure 5: Ablation study on the Geospatial Semantic Alignment (GSA) loss. The inclusion of GSA [PITH_FULL_IMAGE:figures/full_fig_p007_5.png]
Figure 6
Figure 6. Figure 6: Ablation study on the Geospatial Se￾mantic Alignment (GSA) loss. The inclusion of GSA loss improves structural fidelity [PITH_FULL_IMAGE:figures/full_fig_p008_6.png]
Figure 7
Figure 7. Figure 7: Qualitative comparison on practical downstream tasks. (a) Cloud removal: GeoCore-9B [PITH_FULL_IMAGE:figures/full_fig_p009_7.png]
Figure 8
Figure 8. Figure 8: Qualitative evaluation of VAE reconstruction on Earth Observation data. Top row: Original [PITH_FULL_IMAGE:figures/full_fig_p011_8.png]
Figure 9
Figure 9. Figure 9: Qualitative comparison of text-conditioned generation. Given various text descriptions, [PITH_FULL_IMAGE:figures/full_fig_p015_9.png]
Figure 10
Figure 10. Figure 10: Qualitative comparison across varying ground sample distances (GSD). GeoCore-9B [PITH_FULL_IMAGE:figures/full_fig_p016_10.png]
Figure 11
Figure 11. Figure 11: Qualitative comparison across varying ground sample distances (GSD). GeoCore-9B [PITH_FULL_IMAGE:figures/full_fig_p017_11.png]
Figure 12
Figure 12. Figure 12: Qualitative comparison across varying ground sample distances (GSD). GeoCore-9B [PITH_FULL_IMAGE:figures/full_fig_p018_12.png]
Figure 13
Figure 13. Figure 13: Qualitative comparison on practical downstream tasks (cloud removal). GeoCore-9B [PITH_FULL_IMAGE:figures/full_fig_p019_13.png]
Figure 14
Figure 14. Figure 14: Qualitative comparison on practical downstream tasks (SAR-to-optical image translation). [PITH_FULL_IMAGE:figures/full_fig_p019_14.png]

Review history (2 revisions) →

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

58 extracted references · 29 canonical work pages

  1. [1]

    Hongruixuan Chen, Jian Song, Olivier Dietrich, Clifford Broni-Bediako, Weihao Xuan, Junjue Wang, Xinlei Shao, Yimin Wei, Junshi Xia, Cuiling Lan, Konrad Schindler, and Naoto Yokoya. Bright: A globally distributed multimodal building damage assessment dataset with very-high- resolution for all-weather disaster response.Earth System Science Data, 17(11):621...

    work page 2025

  2. [2]

    C-diffset: Leveraging latent diffusion for sar-to-eo image translation with confidence-guided reliable object generation.arXiv preprint arXiv:2411.10788, 2024

    Jeonghyeok Do, Jaehyup Lee, and Munchurl Kim. C-diffset: Leveraging latent diffusion for sar-to-eo image translation with confidence-guided reliable object generation.arXiv preprint arXiv:2411.10788, 2024

    arXiv 2024

  3. [3]

    Uncrtaints: Uncertainty quantification for cloud removal in optical satellite time series

    Patrick Ebel, Vivien Sainte Fare Garnot, Michael Schmitt, Jan Dirk Wegner, and Xiao Xiang Zhu. Uncrtaints: Uncertainty quantification for cloud removal in optical satellite time series. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pages 2086–2096, 2023

    work page 2023

  4. [4]

    Sen12ms-cr-ts: A remote-sensing data set for multimodal multitemporal cloud removal.IEEE Transactions on Geoscience and Remote Sensing, 60:1–14, 2022

    Patrick Ebel, Yajin Xu, Michael Schmitt, and Xiao Xiang Zhu. Sen12ms-cr-ts: A remote-sensing data set for multimodal multitemporal cloud removal.IEEE Transactions on Geoscience and Remote Sensing, 60:1–14, 2022

    work page 2022

  5. [5]

    Filmy cloud removal on satellite imagery with multispectral conditional generative adversarial nets

    Kenji Enomoto, Ken Sakurada, Weimin Wang, Hiroshi Fukui, Masashi Matsuoka, Ryosuke Nakamura, and Nobuo Kawaguchi. Filmy cloud removal on satellite imagery with multispectral conditional generative adversarial nets. InProceedings of the IEEE conference on computer vision and pattern recognition workshops, pages 48–56, 2017

    work page 2017

  6. [6]

    Scaling rectified flow trans- formers for high-resolution image synthesis

    Patrick Esser, Sumith Kulal, Andreas Blattmann, Rahim Entezari, Jonas Müller, Harry Saini, Yam Levi, Dominik Lorenz, Axel Sauer, Frederic Boesel, et al. Scaling rectified flow trans- formers for high-resolution image synthesis. InForty-first international conference on machine learning, 2024

    2024

  7. [7]

    Patrick Helber, Benjamin Bischke, Andreas Dengel, and Damian Borth. Eurosat: A novel dataset and deep learning benchmark for land use and land cover classification.IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing, 12(7):2217–2226, 2019

    2019

  8. [8]

    Classifier-free diffusion guidance.arXiv preprint arXiv:2207.12598, 2022

    Jonathan Ho and Tim Salimans. Classifier-free diffusion guidance.arXiv preprint arXiv:2207.12598, 2022

    arXiv 2022

Show all 58 references
  1. [9]

    Lora: Low-rank adaptation of large language models.Iclr, 1(2):3, 2022

    Edward J Hu, Yelong Shen, Phillip Wallis, Zeyuan Allen-Zhu, Yuanzhi Li, Shean Wang, Liang Wang, Weizhu Chen, et al. Lora: Low-rank adaptation of large language models.Iclr, 1(2):3, 2022

    2022

  2. [10]

    Ctgan: Cloud transformer generative adversarial network

    Gi-Luen Huang and Pei-Yuan Wu. Ctgan: Cloud transformer generative adversarial network. In 2022 IEEE International Conference on Image Processing (ICIP), pages 511–515. IEEE, 2022

    2022

  3. [11]

    The qxs-saropt dataset for deep learning in sar-optical data fusion.arXiv preprint arXiv:2103.08259, 2021

    Meiyu Huang, Yao Xu, Lixin Qian, Weili Shi, Yaqin Zhang, Wei Bao, Nan Wang, Xuejiao Liu, and Xueshuang Xiang. The qxs-saropt dataset for deep learning in sar-optical data fusion.arXiv preprint arXiv:2103.08259, 2021

    2021 arXiv

  4. [12]

    Image-to-image translation with conditional adversarial networks

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

    2017

  5. [13]

    Terramind: Large-scale generative multimodality for earth observation

    Johannes Jakubik, Felix Yang, Benedikt Blumenstiel, Erik Scheurer, Rocco Sedona, Stefano Maurogiovanni, Jente Bosmans, Nikolaos Dionelis, Valerio Marsocci, Niklas Kopp, et al. Terramind: Large-scale generative multimodality for earth observation. InProceedings of the IEEE/CVF ...

    2025

  6. [14]

    Can generative geospatial diffusion models excel as discriminative geospatial foundation mod- els? InProceedings of the IEEE/CVF International Conference on Computer Vision, 2025

    Yuru Jia, Valerio Marsocci, Ziyang Gong, Xue Yang, Maarten Vergauwen, and Andrea Nascetti. Can generative geospatial diffusion models excel as discriminative geospatial foundation mod- els? InProceedings of the IEEE/CVF International Conference on Computer Vision, 2025. 20

    2025

  7. [15]

    Denoising diffusion probabilistic feature-based network for cloud removal in sentinel-2 imagery.Remote Sensing, 15(9):2217, 2023

    Ran Jing, Fuzhou Duan, Fengxian Lu, Miao Zhang, and Wenji Zhao. Denoising diffusion probabilistic feature-based network for cloud removal in sentinel-2 imagery.Remote Sensing, 15(9):2217, 2023

    2023

  8. [16]

    Lobell, and Stefano Ermon

    Samar Khanna, Patrick Liu, Linqi Zhou, Chenlin Meng, Robin Rombach, Marshall Burke, David B. Lobell, and Stefano Ermon. Diffusionsat: A generative foundation model for satellite imagery. InThe Twelfth International Conference on Learning Representations, 2024

    2024

  9. [17]

    Conditional brownian bridge diffusion model for vhr sar to optical image translation.IEEE Geoscience and Remote Sensing Letters, 2025

    Seon-Hoon Kim and Daewon Chung. Conditional brownian bridge diffusion model for vhr sar to optical image translation.IEEE Geoscience and Remote Sensing Letters, 2025

    2025

  10. [18]

    Black Forest Labs, Stephen Batifol, Andreas Blattmann, Frederic Boesel, Saksham Consul, Cyril Diagne, Tim Dockhorn, Jack English, Zion English, Patrick Esser, et al. Flux. 1 kontext: Flow matching for in-context image generation and editing in latent space.arXiv preprint arXiv...

    2025 arXiv

  11. [19]

    Cfca-set: Coarse-to-fine context-aware sar-to-eo translation with auxiliary learning of sar-to-nir translation.IEEE Transactions on Geoscience and Remote Sensing, 61:1–18, 2023

    Jaehyup Lee, Hyebin Cho, Doochun Seo, Hyun-Ho Kim, Jaeheon Jeong, and Munchurl Kim. Cfca-set: Coarse-to-fine context-aware sar-to-eo translation with auxiliary learning of sar-to-nir translation.IEEE Transactions on Geoscience and Remote Sensing, 61:1–18, 2023

    2023

  12. [20]

    Repa-e: Unlocking vae for end-to-end tuning of latent diffusion transformers

    Xingjian Leng, Jaskirat Singh, Yunzhong Hou, Zhenchang Xing, Saining Xie, and Liang Zheng. Repa-e: Unlocking vae for end-to-end tuning of latent diffusion transformers. InProceedings of the IEEE/CVF International Conference on Computer Vision, pages 18262–18272, 2025

    2025

  13. [21]

    Bbdm: Image-to-image translation with brownian bridge diffusion models

    Bo Li, Kaitao Xue, Bin Liu, and Yu-Kun Lai. Bbdm: Image-to-image translation with brownian bridge diffusion models. InProceedings of the IEEE/CVF conference on computer vision and pattern Recognition, pages 1952–1961, 2023

    1952

  14. [22]

    Flow matching for generative modeling.arXiv preprint arXiv:2210.02747, 2022

    Yaron Lipman, Ricky TQ Chen, Heli Ben-Hamu, Maximilian Nickel, and Matt Le. Flow matching for generative modeling.arXiv preprint arXiv:2210.02747, 2022

    2022 arXiv

  15. [23]

    Text2earth: Unlocking text-driven remote sensing image generation with a global-scale dataset and a foundation model

    Chenyang Liu, Keyan Chen, Rui Zhao, Zhengxia Zou, and Zhenwei Shi. Text2earth: Unlocking text-driven remote sensing image generation with a global-scale dataset and a foundation model. IEEE Geoscience and Remote Sensing Magazine, 2025

    2025

  16. [24]

    Sarmae: Masked autoencoder for sar representation learning

    Danxu Liu, Di Wang, Hebaixu Wang, Haoyang Chen, Wentao Jiang, Yilin Cheng, Haonan Guo, Wei Cui, and Jing Zhang. Sarmae: Masked autoencoder for sar representation learning. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 2026

    2026

  17. [25]

    Flow straight and fast: Learning to generate and transfer data with rectified flow.arXiv preprint arXiv:2209.03003, 2022

    Xingchao Liu, Chengyue Gong, and Qiang Liu. Flow straight and fast: Learning to generate and transfer data with rectified flow.arXiv preprint arXiv:2209.03003, 2022

    2022 arXiv

  18. [26]

    Effective cloud removal for remote sensing images by an improved mean-reverting denoising model with elucidated design space

    Yi Liu, Wengen Li, Jihong Guan, Shuigeng Zhou, and Yichao Zhang. Effective cloud removal for remote sensing images by an improved mean-reverting denoising model with elucidated design space. InProceedings of the Computer Vision and Pattern Recognition Conference, pages 17851–1...

    2025

  19. [27]

    Diffusion models meet remote sensing: Principles, methods, and perspectives.IEEE Transactions on Geoscience and Remote Sensing, 62:1–22, 2024

    Yidan Liu, Jun Yue, Shaobo Xia, Pedram Ghamisi, Weiying Xie, and Leyuan Fang. Diffusion models meet remote sensing: Principles, methods, and perspectives.IEEE Transactions on Geoscience and Remote Sensing, 62:1–22, 2024

    2024

  20. [28]

    Exploring models and data for remote sensing image caption generation.IEEE Transactions on Geoscience and Remote Sensing, 56(4):2183–2195, 2017

    Xiaoqiang Lu, Binqiang Wang, Xiangtao Zheng, and Xuelong Li. Exploring models and data for remote sensing image caption generation.IEEE Transactions on Geoscience and Remote Sensing, 56(4):2183–2195, 2017

    2017

  21. [29]

    Cloud removal in sentinel-2 imagery using a deep residual neural network and sar-optical data fusion.ISPRS Journal of Photogrammetry and Remote Sensing, 166:333–346, 2020

    Andrea Meraner, Patrick Ebel, Xiao Xiang Zhu, and Michael Schmitt. Cloud removal in sentinel-2 imagery using a deep residual neural network and sar-optical data fusion.ISPRS Journal of Photogrammetry and Remote Sensing, 166:333–346, 2020

    2020

  22. [30]

    Scalable diffusion models with transformers

    William Peebles and Saining Xie. Scalable diffusion models with transformers. InProceedings of the IEEE/CVF international conference on computer vision, pages 4195–4205, 2023. 21

    2023

  23. [31]

    Learning transferable visual models from natural language supervision

    Alec Radford, Jong Wook Kim, Chris Hallacy, Aditya Ramesh, Gabriel Goh, Sandhini Agarwal, Girish Sastry, Amanda Askell, Pamela Mishkin, Jack Clark, et al. Learning transferable visual models from natural language supervision. InInternational conference on machine learning, pag...

    2021

  24. [32]

    Exploring the limits of transfer learning with a unified text-to-text transformer.Journal of machine learning research, 21(140):1–67, 2020

    Colin Raffel, Noam Shazeer, Adam Roberts, Katherine Lee, Sharan Narang, Michael Matena, Yanqi Zhou, Wei Li, and Peter J Liu. Exploring the limits of transfer learning with a unified text-to-text transformer.Journal of machine learning research, 21(140):1–67, 2020

    2020

  25. [33]

    Zero: Memory optimiza- tions toward training trillion parameter models

    Samyam Rajbhandari, Jeff Rasley, Olatunji Ruwase, and Yuxiong He. Zero: Memory optimiza- tions toward training trillion parameter models. InSC20: international conference for high performance computing, networking, storage and analysis, pages 1–16. IEEE, 2020

    2020

  26. [34]

    Zero-shot text-to-image generation

    Aditya Ramesh, Mikhail Pavlov, Gabriel Goh, Scott Gray, Chelsea V oss, Alec Radford, Mark Chen, and Ilya Sutskever. Zero-shot text-to-image generation. InInternational conference on machine learning, pages 8821–8831. Pmlr, 2021

    2021

  27. [35]

    High- resolution image synthesis with latent diffusion models

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

    2022

  28. [36]

    Dae- gan: Dynamic aspect-aware gan for text-to-image synthesis

    Shulan Ruan, Yong Zhang, Kun Zhang, Yanbo Fan, Fan Tang, Qi Liu, and Enhong Chen. Dae- gan: Dynamic aspect-aware gan for text-to-image synthesis. InProceedings of the IEEE/CVF international conference on computer vision, pages 13960–13969, 2021

    2021

  29. [37]

    Cloud removal from satellite images using spatiotemporal generator networks

    Vishnu Sarukkai, Anirudh Jain, Burak Uzkent, and Stefano Ermon. Cloud removal from satellite images using spatiotemporal generator networks. InProceedings of the IEEE/CVF winter conference on applications of computer vision, pages 1796–1805, 2020

    2020

  30. [38]

    Rsdiff: Remote sensing image generation from text using diffusion model.Neural Computing and Applications, 36(36):23103–23111, 2024

    Ahmad Sebaq and Mohamed ElHelw. Rsdiff: Remote sensing image generation from text using diffusion model.Neural Computing and Applications, 36(36):23103–23111, 2024

    2024

  31. [39]

    Oriane Siméoni, Huy V V o, Maximilian Seitzer, Federico Baldassarre, Maxime Oquab, Cijo Jose, Vasil Khalidov, Marc Szafraniec, Seungeun Yi, Michaël Ramamonjisoa, et al. Dinov3. arXiv preprint arXiv:2508.10104, 2025

    2025 arXiv

  32. [40]

    Roformer: Enhanced transformer with rotary position embedding.Neurocomputing, 568:127063, 2024

    Jianlin Su, Murtadha Ahmed, Yu Lu, Shengfeng Pan, Wen Bo, and Yunfeng Liu. Roformer: Enhanced transformer with rotary position embedding.Neurocomputing, 568:127063, 2024

    2024

  33. [41]

    Crs-diff: Controllable remote sensing image generation with diffusion model.IEEE Transac- tions on Geoscience and Remote Sensing, 62:1–14, 2024

    Datao Tang, Xiangyong Cao, Xingsong Hou, Zhongyuan Jiang, Junmin Liu, and Deyu Meng. Crs-diff: Controllable remote sensing image generation with diffusion model.IEEE Transac- tions on Geoscience and Remote Sensing, 62:1–14, 2024

    2024

  34. [42]

    Df-gan: A simple and effective baseline for text-to-image synthesis

    Ming Tao, Hao Tang, Fei Wu, Xiao-Yuan Jing, Bing-Kun Bao, and Changsheng Xu. Df-gan: A simple and effective baseline for text-to-image synthesis. InProceedings of the IEEE/CVF conference on computer vision and pattern recognition, pages 16515–16525, 2022

    2022

  35. [43]

    Loveda: A remote sensing land-cover dataset for domain adaptive semantic segmentation

    Junjue Wang, Zhuo Zheng, Ailong Ma, Xiaoyan Lu, and Yanfei Zhong. Loveda: A remote sensing land-cover dataset for domain adaptive semantic segmentation. InAdvances in Neural Information Processing Systems Datasets and Benchmarks Track, volume 1, 2021

    2021

  36. [44]

    Stegogan: Leveraging steganography for non-bijective image-to-image translation

    Sidi Wu, Yizi Chen, Samuel Mermet, Lorenz Hurni, Konrad Schindler, Nicolas Gonthier, and Loic Landrieu. Stegogan: Leveraging steganography for non-bijective image-to-image translation. InProceedings of the IEEE/CVF conference on computer vision and pattern recognition, pages 7...

    2024

  37. [45]

    Attngan: Fine-grained text to image generation with attentional generative adversarial networks

    Tao Xu, Pengchuan Zhang, Qiuyuan Huang, Han Zhang, Zhe Gan, Xiaolei Huang, and Xiaodong He. Attngan: Fine-grained text to image generation with attentional generative adversarial networks. InProceedings of the IEEE conference on computer vision and pattern recognition, pages 1...

    2018

  38. [46]

    Txt2img- mhn: Remote sensing image generation from text using modern hopfield networks.IEEE Transactions on Image Processing, 32:5737–5750, 2023

    Yonghao Xu, Weikang Yu, Pedram Ghamisi, Michael Kopp, and Sepp Hochreiter. Txt2img- mhn: Remote sensing image generation from text using modern hopfield networks.IEEE Transactions on Image Processing, 32:5737–5750, 2023. 22

    2023

  39. [47]

    Reconstruction vs

    Jingfeng Yao, Bin Yang, and Xinggang Wang. Reconstruction vs. generation: Taming optimiza- tion dilemma in latent diffusion models. InProceedings of the Computer Vision and Pattern Recognition Conference, pages 15703–15712, 2025

    2025

  40. [48]

    Deterministic guid- ance diffusion model for probabilistic weather forecasting.arXiv preprint arXiv:2312.02819, 2023

    Donggeun Yoon, Minseok Seo, Doyi Kim, Yeji Choi, and Donghyeon Cho. Deterministic guid- ance diffusion model for probabilistic weather forecasting.arXiv preprint arXiv:2312.02819, 2023

    2023 arXiv

  41. [49]

    Transformer-based synthetic-to-measured sar image translation via learning of representational features.IEEE Transactions on Geoscience and Remote Sensing, 61:1–18, 2023

    Geunhyuk Youk and Munchurl Kim. Transformer-based synthetic-to-measured sar image translation via learning of representational features.IEEE Transactions on Geoscience and Remote Sensing, 61:1–18, 2023

    2023

  42. [50]

    Representation alignment for generation: Training diffusion transformers is easier than you think.arXiv preprint arXiv:2410.06940, 2024

    Sihyun Yu, Sangkyung Kwak, Huiwon Jang, Jongheon Jeong, Jonathan Huang, Jinwoo Shin, and Saining Xie. Representation alignment for generation: Training diffusion transformers is easier than you think.arXiv preprint arXiv:2410.06940, 2024

    arXiv

  43. [51]

    Metaearth: A generative foundation model for global-scale remote sensing image generation.IEEE Transactions on Pattern Analysis and Machine Intelligence, 47(3):1764–1781, 2024

    Zhiping Yu, Chenyang Liu, Liqin Liu, Zhenwei Shi, and Zhengxia Zou. Metaearth: A generative foundation model for global-scale remote sensing image generation.IEEE Transactions on Pattern Analysis and Machine Intelligence, 47(3):1764–1781, 2024

    2024

  44. [52]

    Adding conditional control to text-to-image diffusion models

    Lvmin Zhang, Anyi Rao, and Maneesh Agrawala. Adding conditional control to text-to-image diffusion models. InProceedings of the IEEE/CVF international conference on computer vision, pages 3836–3847, 2023

    2023

  45. [53]

    Text-to-remote-sensing-image generation with structured generative adversarial networks.IEEE Geoscience and Remote Sensing Letters, 19:1–5, 2021

    Rui Zhao and Zhenwei Shi. Text-to-remote-sensing-image generation with structured generative adversarial networks.IEEE Geoscience and Remote Sensing Letters, 19:1–5, 2021

    2021

  46. [54]

    Uni-controlnet: All-in-one control to text-to-image diffusion models

    Shihao Zhao, Dongdong Chen, Yen-Chun Chen, Jianmin Bao, Shaozhe Hao, Lu Yuan, and Kwan-Yee K Wong. Uni-controlnet: All-in-one control to text-to-image diffusion models. Advances in neural information processing systems, 36:11127–11150, 2023

    2023

  47. [55]

    Towards language-free training for text-to-image generation

    Yufan Zhou, Ruiyi Zhang, Changyou Chen, Chunyuan Li, Chris Tensmeyer, Tong Yu, Jiuxiang Gu, Jinhui Xu, and Tong Sun. Towards language-free training for text-to-image generation. In Proceedings of the IEEE/CVF conference on computer vision and pattern recognition, pages 17907–1...

    2022

  48. [56]

    Unpaired image-to-image translation using cycle-consistent adversarial networks

    Jun-Yan Zhu, Taesung Park, Phillip Isola, and Alexei A Efros. Unpaired image-to-image translation using cycle-consistent adversarial networks. InProceedings of the IEEE international conference on computer vision, pages 2223–2232, 2017

    2017

  49. [57]

    Pmaa: A progressive multi-scale attention autoencoder model for high-performance cloud removal from multi-temporal satellite imagery.arXiv preprint arXiv:2303.16565, 2023

    Xuechao Zou, Kai Li, Junliang Xing, Pin Tao, and Yachao Cui. Pmaa: A progressive multi-scale attention autoencoder model for high-performance cloud removal from multi-temporal satellite imagery.arXiv preprint arXiv:2303.16565, 2023

    2023 arXiv

  50. [58]

    Diffcr: A fast conditional diffusion framework for cloud removal from optical satellite images.IEEE Transactions on Geoscience and Remote Sensing, 62:1–14, 2024

    Xuechao Zou, Kai Li, Junliang Xing, Yu Zhang, Shiying Wang, Lei Jin, and Pin Tao. Diffcr: A fast conditional diffusion framework for cloud removal from optical satellite images.IEEE Transactions on Geoscience and Remote Sensing, 62:1–14, 2024. 23

    2024

Pith tools

Reviewed August 15, 2026 · model on record in the stance chip above.