Pith. sign in

REVIEW 5 major objections 5 minor 4 cited by

EarthSynth: Generating Informative Earth Observation with Diffusion Models

T0 review · 5 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 EarthSynth generates labeled satellite imagery with a diffusion model and claims that adding it to real data improves open-vocabulary classification, detection, and segmentation over real-data-only training.

desk verdict Real system, useful dataset, believable detection/segmentation gains; scene-classification numbers are inflated by a circular CLIP filter/evaluation design. read the letter →

arxiv 2505.12108 v2 pith:63NLIRBO submitted 2025-05-17 cs.CV cs.AI

Jiancheng Pan , Shiye Lei , Yuqian Fu , Jiahao Li , Yanxing Liu , Yuze Sun , Xiao He , Long Peng
show 2 more authors
Xiaomeng Huang Bo Zhao
This is my paper · ORCID
classification cs.CVcs.AI
keywords EarthSynthremotesensingimageinterpretationdiffusionmodelssyntheticdataaugmentationopen-vocabularysceneclassificationobjectdetectionsemanticsegmentationcounterfactualcomposition
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

EarthSynth addresses the scarcity of labeled remote sensing imagery by replacing task-specific augmentation pipelines with one generative foundation model. The paper constructs a 180,000-sample multi-source training set of image, semantic-mask, and text triples, trains a latent diffusion model on it, and synthesizes labeled images for scene classification, object detection, and semantic segmentation. Its two innovations are Counterfactual Composition, which pastes objects into new backgrounds during training to create plausible scenes that never occurred, and R-Filter, which discards generated samples whose CLIP score falls below 0.4. In open-vocabulary evaluations, synthetic EarthSynth data added to real data improves detection mAP on DOTAv2 from 56.3 to 58.4 and segmentation mIoU on FloodNet from 33.9 to 44.4, beating ControlNet-based synthetic data and real-data-only training. The paper concludes that diverse, category-controlled generation is a practical route to better remote sensing interpretation when labels are scarce.

What carries the argument

Counterfactual Composition (CF-Comp) is the load-bearing training mechanism. In each batch, image-mask-text pairs are screened by three compatibility scores — Image Color Sensitivity in channel space, Mask Overlap Rate in pixel space, and text semantic similarity in semantic space — and compatible pairs are combined by Copy-Paste, pasting one image's foreground objects into the other's background to form an unrealistic logical scene. The model is trained with a global noise-prediction loss plus a local loss that weights error inside the semantic-mask region, which the paper says is necessary for precise layout control. At synthesis time, R-Filter applies a rule-based CLIP-score threshold of 0.4 on the whole image or on the object region, keeping only samples that pass; masks are converted to bounding boxes for detection via contour simplification, completing the image-mask-text-to-label mapping.

What would settle it

Hold the generated image pool fixed, replace the CLIP-ViT-B/32 scorer inside R-Filter with a different pre-trained text-image model (for instance one tuned on remote sensing imagery), keep the threshold at 0.4, and re-measure detection and segmentation; if the mAP and mIoU gains shrink to noise, the filter is selecting for scorer preferences rather than task-relevant content.

Watch

Extended reading notes

Core claim

On the paper's own terms, EarthSynth is a diffusion-based generative foundation model that, given a text prompt and a semantic mask, generates multi-category, cross-satellite Earth observation data with labels already attached. The central claim is that training downstream open-vocabulary models on real data plus EarthSynth-synthesized data outperforms both training on real data alone and training with ControlNet-generated data, across scene classification, object detection, and semantic segmentation. Representative gains claimed are DOTAv2 mAP from 56.3 to 58.4, Potsdam mIoU from 40.6 to 42.7, FloodNet mIoU from 33.9 to 44.4, and FLAIR mIoU from 19.3 to 21.6. The method's rationale is that mixing real-world and unrealistic logical distributions during training increases the model's generative diversity, and that filtering by CLIP score keeps the synthesized samples most useful for downstream tasks.

Load-bearing premise

The load-bearing premise is that a CLIP score above 0.4 marks genuinely informative variety, because the same type of scorer is used both to filter synthetic images and to grade scene-classification success.

Editorial extensions

If this is right

  • One pretrained conditional generator can feed multiple interpretation tasks, so downstream users can avoid training a separate synthesis model for classification, detection, and segmentation.
  • Synthetic data helps most at moderate scale: the DOTAv2 ablation shows mAP peaking at 128 generated samples per class and declining with more, so synthetic volume should be budgeted rather than maximized.
  • The CLIP-score filter is a reusable data sieve: because R-Filter retains samples whose overall or object-level score exceeds 0.4, organizations can curate synthetic training pools without manual inspection.
  • Adding EarthSynth data to real data improves open-vocabulary few-shot detection, for example raising 10-shot DIOR mAP from 37.5 with real data only to 40.7 with synthetic data included.

Reading between the lines

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

  • The paper does not run a control in which R-Filter's CLIP scorer is replaced by a different text-image model; if detection and segmentation gains vanish under that substitution, the filter's benefit is scorer-specific rather than a general property of the generated images.
  • The scene-classification results should be read with caution because the same CLIP family both filters the synthetic data and measures classification accuracy; the detection and segmentation numbers, which use different downstream models, are the cleaner evidence for the paper's claim.
  • The paper's own limitations state that multispectral generalization is unproven and training cost is high; an extension would train and evaluate on multispectral bands, where the cross-sensor premise could be stress-tested.
  • A natural next experiment is a self-training loop in which EarthSynth is retrained on its own R-Filtered outputs; the diversity-by-composition mechanism may delay, but not necessarily prevent, the model-collapse effects seen in prior self-consuming generation work.
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

5 major / 5 minor

Summary. This manuscript introduces EarthSynth, a conditional diffusion model for remote sensing image synthesis that generates images together with aligned semantic masks and text prompts, trained on a new multi-source dataset EarthSynth-180K. The methodological contributions are a Counterfactual Composition (CF-Comp) training strategy with channel, pixel, and semantic similarity criteria and a local mask loss, plus a CLIP-score-based R-Filter for selecting synthetic samples. The authors evaluate synthetic data on open-vocabulary scene classification with CLIP, object detection with GroundingDINO, and semantic segmentation with GSNet, reporting improvements over real-data-only training and over ControlNet-generated data, e.g., DOTAv2 mAP 56.3 to 58.4 and FloodNet mIoU 33.9 to 44.4.

Significance. If the results hold, EarthSynth is a useful step toward multi-task synthetic data generation for remote sensing. The detection and segmentation gains are credible evidence because they use independent downstream models and real test sets, and the ablation in Table 5 supports the local-loss and CF-Comp contributions, modulo the threshold inconsistency discussed below. The EarthSynth-180K dataset, with aligned image, mask, and text triplets, is a practical asset, and the paper explicitly reports training cost and limitations. However, the scene-classification protocol is circular with respect to R-Filter, and several result statements are internally inconsistent; these issues must be fixed before the classification and filtering claims can be accepted.

major comments (5)
  1. [Sec. 6.1 (Table 1), Algorithm 2, R-Filter] The scene-classification evaluation is circular with respect to R-Filter. In Algorithm 2, a generated sample is retained only if the CLIP score against the condition text exceeds S0=0.4, and Section 6.1 then evaluates classification on those retained images with CLIP-ViT-B/32 using the same text prompts (Table 1; Appendix Table 11 has the same structure). Since the evaluation model is exactly the model used for selection, high classification accuracy can reflect selection by the evaluation criterion rather than the informativeness of the generated data for an independent task. No classification result is reported on unfiltered synthetic data or on data filtered with an independent model, so the R-Filter contribution to scene classification and the Table 1 comparison against ControlNet are not interpretable. Please add an unfiltered baseline or use a different open-vocabulary evaluator (e.g., a different CLIP backbone or a non-CLIP model) for the scene-classification metric.
  2. [Sec. 4 (Algorithm 1), Sec. 6 (Experiment Setup)] The CF-Comp condition in Algorithm 1 cannot be satisfied with the stated hyperparameters. The algorithm applies Copy-Paste only when α, β, η > α0, β0, η0, but the setup states α0=1, and ICS is defined in Eq. (4) to take values in {0,1}. Consequently α > 1 never holds, the composition branch never executes, and the CF-Comp improvement reported in Table 5 (57.4 to 57.9 mAP) is not reproducible from the paper. Please state the intended threshold (e.g., α0=0 or a non-strict comparison) and report the actual number of composed samples used.
  3. [Sec. 6.1 (Table 3)] The few-shot detection summary in Section 6.1 is contradicted by Table 3. The text states that ETS 'consistently outperforms close-source baselines across all shot settings,' but in the 1-shot block ETS with real data (12.7 mAP) is below DE-ViT (14.7) and CD-ViTO (17.8), and ETS with EarthSynth data (13.9) remains below both. Please revise the claim to reflect the settings and the variant (Real vs. Real+Synth) for which gains actually occur.
  4. [Appendix A.4.2 (Table 10)] The GPT-4 scoring paragraph in Appendix A.4.2 contradicts Table 10. The text says 'ControlNet achieved a higher overall average score of 53.07,' while Table 10 lists EarthSynth at 53.07 and ControlNet at 51.64, and the following sentence correctly concludes that EarthSynth is better. This contradiction must be corrected, and the discussion should be reconciled with the visual-quality and semantic-richness subscores in the same table.
  5. [Sec. 1 (Contributions), Sec. 2 (Related Work), Appendix Table 9] The novelty claim for EarthSynth-180K is overstated. The contributions state that EarthSynth-180K is 'the first large-scale remote sensing dataset for diffusion training,' but the related work cites Text2Earth [19], which is described as using a global-scale dataset (Git-10M) for text-driven remote sensing image generation, and Appendix Table 9 lists Text2Earth trained on Git-10M. Please qualify the claim and situate EarthSynth-180K relative to existing large-scale remote sensing generation datasets.
minor comments (5)
  1. [Tables 1, 3, 4] The comparative tables do not report variance or confidence intervals; Table 1 states that results are averaged over three runs, but no standard deviation is given, and detection and segmentation appear to be single runs.
  2. [Algorithm 2] Algorithm 2 uses '|' in the filtering condition 'Score_CLIP(x)>S0 | Score_CLIP(xobj)>S0'; this should be a logical OR, and the composition of the overall versus object/background scores should be defined precisely.
  3. [Sec. 6.3 (Figure 5)] Figure 5 reports FID 335.0 for EarthSynth versus 310.9 for ControlNet; since higher FID usually indicates worse fidelity to the reference distribution, the sentence claiming 'higher-quality images' should be reworded to distinguish deliberate distributional shift from visual quality.
  4. [Throughout] There are several typos, including 'The the generative distribution' (Sec. 3), 'downtown models' and 'downtown evaluation' (Secs. 5 and 6), and 'close-source' (Sec. 6.1).
  5. [Appendix A.4.1 (Table 9)] Table 9 compares FID and CLIP scores across methods, but values marked '*' are quoted from different original papers with different reference distributions; the table should state that these numbers are not directly comparable.

Circularity Check

1 steps flagged · score 6.0 of 10

Scene-classification gains are partially circular: R-Filter selects images by CLIP score and the same CLIP model measures classification accuracy; detection and segmentation remain independent.

  1. self definitional [Section 5 'Rule-based Filtering', Algorithm 2; Section 6 'Experiment Setup'; Section 6.1 'Scene Classification', Table 1]
    "Rule-based Filtering. We propose R-Filter, a rule-based method that uses CLIP scores to evaluate {image x, object xobj, background xbg} triplets ... we retain samples with high overall or object-specific scores for training downtown models by setting the CLIP score threshold S0. ... if Score CLIP(x)>S 0|Score CLIP(xobj)>S 0 then Add generated sample to dataset ... Table 1 presents the comparative results of CLIP-based scene classification regarding Top-1 and Top-5 accuracy on RSICD and DIOR datasets."

    R-Filter retains synthetic images whose CLIP score against the text prompt exceeds S0=0.4, using CLIP-ViT-B/32. The scene-classification evaluation then uses the same CLIP model to measure top-1/top-5 accuracy of the retained images against the same text prompts. Retained images are therefore selected for high values of the exact quantity being reported as classification accuracy, so the Table 1 numbers are forced upward relative to unfiltered generator output. The marginal benefit of R-Filter for classification in Table 11 (49.07 to 50.47) is likewise a selection artifact unless evaluated on unfiltered or independently filtered data, which the paper does not provide.

full rationale

The paper's main claim is that EarthSynth synthetic data improves open-vocabulary scene classification, object detection, and semantic segmentation. The detection and segmentation evidence is independent: GroundingDINO and GSNet are external models evaluated on real test sets, and the gains over real-only training and ControlNet are not defined by the synthetic-data filter. The circularity is confined to scene classification. There, R-Filter explicitly keeps samples with CLIP score above a threshold, and the reported metric is CLIP-based classification accuracy on those same samples. This is a self-definitional selection step: the filter optimizes the evaluation score, so the classification numbers cannot be interpreted as an unbiased measure of generation quality or of R-Filter's contribution. The paper does not report unfiltered classification numbers for the Table 1 protocol, leaving the size of the artifact unquantified. I do not find load-bearing circularity in the other components: the CF-Comp training objective, the mask-conditioned generation, and the downstream detection/segmentation evaluations are externally grounded. The self-citations in the paper, such as the ETS detector and LAE-1M dataset, are not used to justify the central derivation chain. Overall, one prediction channel reduces by construction while the majority of the empirical support remains independent, giving a partial-circularity score of 6.

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

The method relies on several hand-set thresholds (S0=0.4, gamma=10, CF-Comp thresholds) and on unvalidated modeling assumptions about Gaussian object/background statistics and CLIP score as an informativeness proxy. No new physical or conceptual entity is introduced.

free parameters (4)
  • CLIP score threshold S0 = 0.4
    Used in R-Filter to discard generated samples. The value is set without a sensitivity analysis or principled selection; it directly affects all downstream results.
  • Local loss weight gamma = 10
    Balances global and mask-local noise-prediction losses in Eq (9). Chosen based on qualitative convergence (Appendix A.4.3), no quantitative sweep.
  • CF-Comp thresholds (s0, alpha0, beta0, eta0) = 150, 1, 0.02, 0.6
    Control the three-dimensional batch-sample selection in Algorithm 1. Hand-set; alpha0=1 is unsatisfiable given ICS returns 0 or 1, indicating a likely typo for 0.
  • Synthetic images per category = 256 stated; 128 optimal per Figure 4
    The main experiments state 256 generated images per category, but the sample-size ablation shows best mAP at 128 with decline thereafter, creating an ambiguity in the reported results.
assumptions (5)
  • domain assumption Stable Diffusion v1-5 pretrained weights transfer to remote sensing imagery.
    EarthSynth is initialized from Stable Diffusion v1-5 and fine-tuned on EarthSynth-180K; the transferability of natural-image representations to satellite images is assumed (Section 6, Experiment Setup).
  • domain assumption Satellite image content decomposes as x = f(x_obj, x_bg, x_noise) with x_noise independent of label y.
    Feature Decomposition in Section 3, Eq (2); used to justify why counterfactual composition can improve label-informative generation.
  • ad hoc to paper Object and background pixels are approximately Gaussian distributed.
    Eq (7) assumes Gaussian object and background distributions to derive the mixed mean and variance in Eq (8); no empirical validation is provided.
  • ad hoc to paper CLIP score is a valid proxy for the informativeness of synthetic samples for downstream tasks.
    R-Filter (Section 5) retains samples with CLIP score greater than 0.4; whether this selects informative data or simply easy samples is not established.
  • domain assumption Counterfactual composition is valid only when channel counts match, mask overlap is low, and text semantics are similar.
    Algorithm 1 uses ICS, MOR, and TSS thresholds to decide copy-paste pairs; these criteria are asserted rather than derived.

how reviews work

0 comments
Cite this review

Pith. "Pith review of EarthSynth: Generating Informative Earth Observation with Diffusion Models." pith.science (2026). https://pith.science/paper/63NLIRBO

@misc{pith2026250512108,
  author       = {Pith},
  title        = {Pith review of: EarthSynth: Generating Informative Earth Observation with Diffusion Models},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/63NLIRBO}},
  note         = {Machine review of arXiv:2505.12108}
}
read the original abstract

Remote sensing image (RSI) interpretation typically faces challenges due to the scarcity of labeled data, which limits the performance of RSI interpretation tasks. To tackle this challenge, we propose EarthSynth, a diffusion-based generative foundation model that enables synthesizing multi-category, cross-satellite labeled Earth observation for downstream RSI interpretation tasks. To the best of our knowledge, EarthSynth is the first to explore multi-task generation for remote sensing, tackling the challenge of limited generalization in task-oriented synthesis for RSI interpretation. EarthSynth, trained on the EarthSynth-180K dataset, employs the Counterfactual Composition training strategy with a three-dimensional batch-sample selection mechanism to improve training data diversity and enhance category control. Furthermore, a rule-based method of R-Filter is proposed to filter more informative synthetic data for downstream tasks. We evaluate our EarthSynth on scene classification, object detection, and semantic segmentation in open-world scenarios. There are significant improvements in open-vocabulary understanding tasks, offering a practical solution for advancing RSI interpretation.

Figures

Figures reproduced from arXiv: 2505.12108 by the authors.

Figure 1
Figure 1. A diffusion-based generative foundation model, EarthSynth, pretrained on multi-source and [PITH_FULL_IMAGE:figures/full_fig_p001_1.png] view at source ↗
Figure 2
Figure 2. EarthSynth is trained with CF-Comp training strategy on real and unrealistic data distribu [PITH_FULL_IMAGE:figures/full_fig_p003_2.png] view at source ↗
Figure 3
Figure 3. Left: Copy-Paste used in CF-Comp Strategy. Right: CLIP-based rule filtering retains [PITH_FULL_IMAGE:figures/full_fig_p004_3.png] view at source ↗
Figures from the paper (12 more)
Figure 4
Figure 4. Figure 4: Effect of samples per class on DOTAv2 dataset. Method mAP *only real 56.3 baseline - + Llocal 56.5 + Llocal + R-Filter 57.4 + CF-Comp - + CF-Comp + Llocal 57.9 + CF-Comp + Llocal + R-Filter 58.4 [PITH_FULL_IMAGE:figures/full_fig_p009_4.png]
Figure 5
Figure 5. Figure 5: Left: Visualization of synthesis satellite images on DOTAv2 dataset. Right: EarthSynth can [PITH_FULL_IMAGE:figures/full_fig_p010_5.png]
Figure 6
Figure 6. Figure 6: Embedding visualizations for some categories on the EarthSynth-180K dataset. [PITH_FULL_IMAGE:figures/full_fig_p010_6.png]
Figure 7
Figure 7. Figure 7: Category distribution of the EarthSynth-180K. [PITH_FULL_IMAGE:figures/full_fig_p016_7.png]
Figure 8
Figure 8. Figure 8: A category-augmentation strategy to construct multiple object-background pairs. [PITH_FULL_IMAGE:figures/full_fig_p017_8.png]
Figure 9
Figure 9. Figure 9: Some images of the EarthSynth-180K dataset are derived from SAR imagery. SAMRS. SAMRS dataset is a large-scale re￾mote sensing segmentation dataset developed us￾ing the Segment Anything Model [69]. It lever￾ages existing remote sensing object detection datasets to gene…
Figure 10
Figure 10. Figure 10: EarthSynth over time-step training process. [PITH_FULL_IMAGE:figures/full_fig_p021_10.png]
Figure 11
Figure 11. Figure 11: EarthSynth over time-step training process. [PITH_FULL_IMAGE:figures/full_fig_p022_11.png]
Figure 12
Figure 12. Figure 12: Examples of generated segmentation data. [PITH_FULL_IMAGE:figures/full_fig_p023_12.png]
Figure 13
Figure 13. Figure 13: Examples of generated object detection data. This includes the ability to generate data [PITH_FULL_IMAGE:figures/full_fig_p024_13.png]
Figure 14
Figure 14. Figure 14: Comparison of different generation methods. [PITH_FULL_IMAGE:figures/full_fig_p024_14.png]
Figure 15
Figure 15. Figure 15: The guidance scale affects the CLIP score across two categories. [PITH_FULL_IMAGE:figures/full_fig_p025_15.png]

Discussion (0). Continue with ORCID to comment.

Forward citations

Cited by 4 Pith papers

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

  1. From Synchrony to Sequence: Exo-to-Ego Generation via Interpolation

    cs.CV 2026-04 unverdicted novelty 6.0 of 10

    Interpolating only the video frames between synchronized exo and ego clips already turns discontinuous cross-view generation into continuous sequence modeling and measurably improves diffusion-based Exo2Ego synthesis.

  2. TerraDiT: Point-Conditioned Diffusion Transformer for Satellite Image Synthesis

    cs.CV 2026-03 conditional novelty 6.0 of 10

    GeoDiT is a point-conditioned diffusion transformer that generates satellite imagery from sparse labeled points and claims to beat existing remote sensing generators on FID and SSIM.

  3. Disentangle Identity, Cooperate Emotion: Correlation-Aware Emotional Talking Portrait Generation

    cs.CV 2025-04 conditional novelty 5.0 of 10

    DICE-Talk improves emotional talking-head generation by combining an audio-visual Gaussian emotion prior, a vector-quantized emotion bank, and an auxiliary emotion classifier in a diffusion model.

  4. Control Copy-Paste: Controllable Diffusion-Based Augmentation Method for Remote Sensing Few-Shot Object Detection

    eess.IV 2025-07 conditional novelty 4.0 of 10

    Control Copy-Paste uses AnyDoor's diffusion model to insert few-shot satellite objects into varied contexts, improving DIOR few-shot detection by an average of 10.76% mAP.

Reference graph

Works this paper leans on

71 extracted references · 47 canonical work pages · cited by 4 Pith papers

  1. [19]

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

    C. Liu, K. Chen, R. Zhao, Z. Zou, and Z. Shi, “Text2earth: Unlocking text-driven remote sensing image generation with a global-scale dataset and a foundation model,”arXiv preprint arXiv:2501.00895, 2025

    arXiv 2025

  2. [1]

    Research progress on few-shot learning for remote sensing image interpretation,

    X. Sun, B. Wang, Z. Wang, H. Li, H. Li, and K. Fu, “Research progress on few-shot learning for remote sensing image interpretation,”IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing, vol. 14, pp. 2387–2402, 2021

    work page 2021

  3. [2]

    Addressing class imbalance in remote sensing using deep learning approaches: a systematic literature review,

    S. Sharma and A. Gosain, “Addressing class imbalance in remote sensing using deep learning approaches: a systematic literature review,”Evolutionary Intelligence, vol. 18, no. 1, pp. 1–28, 2025

    work page 2025

  4. [3]

    Data augmentation generative adversarial networks,

    A. Antoniou, A. Storkey, and H. Edwards, “Data augmentation generative adversarial networks,” arXiv preprint arXiv:1711.04340, 2017

    arXiv 2017

  5. [4]

    High-resolution image synthesis with latent diffusion models,

    R. Rombach, A. Blattmann, D. Lorenz, P. Esser, and B. Ommer, “High-resolution image synthesis with latent diffusion models,” inProceedings of the IEEE/CVF conference on computer vision and pattern recognition, pp. 10684–10695, 2022

    2022

  6. [5]

    A comprehensive survey for generative data augmentation,

    Y . Chen, Z. Yan, and Y . Zhu, “A comprehensive survey for generative data augmentation,” Neurocomputing, p. 128167, 2024

    work page 2024

  7. [6]

    Is synthetic data from generative models ready for image recognition?,

    R. He, S. Sun, X. Yu, C. Xue, W. Zhang, P. Torr, S. Bai, and X. Qi, “Is synthetic data from generative models ready for image recognition?,”arXiv preprint arXiv:2210.07574, 2022

    arXiv 2022

  8. [7]

    Effective data augmentation with diffusion models,

    B. Trabucco, K. Doherty, M. Gurinas, and R. Salakhutdinov, “Effective data augmentation with diffusion models,”arXiv preprint arXiv:2302.07944, 2023

    arXiv 2023

Show all 71 references
  1. [8]

    Learning disentangled identifiers for action-customized text-to-image generation,

    S. Huang, B. Gong, Y . Feng, X. Chen, Y . Fu, Y . Liu, and D. Wang, “Learning disentangled identifiers for action-customized text-to-image generation,” inCVPR, 2024

    2024

  2. [9]

    Self-improving diffusion models with synthetic data,

    S. Alemohammad, A. I. Humayun, S. Agarwal, J. Collomosse, and R. Baraniuk, “Self-improving diffusion models with synthetic data,”arXiv preprint arXiv:2408.16333, 2024

    2024 arXiv

  3. [10]

    Self-consuming generative models go MAD,

    S. Alemohammad, J. Casco-Rodriguez, L. Luzi, A. I. Humayun, H. Babaei, D. LeJeune, A. Siahkoohi, and R. Baraniuk, “Self-consuming generative models go MAD,” inThe Twelfth International Conference on Learning Representations, 2024

    2024

  4. [11]

    Diversify your vision datasets with automatic diffusion-based augmentation,

    L. Dunlap, A. Umino, H. Zhang, J. Yang, J. E. Gonzalez, and T. Darrell, “Diversify your vision datasets with automatic diffusion-based augmentation,” inAdvances in Neural Information Processing Systems(A. Oh, T. Naumann, A. Globerson, K. Saenko, M. Hardt, and S. Levine, eds.), 2023

    2023

  5. [12]

    Txt2img-mhn: Remote sensing image generation from text using modern hopfield networks,

    Y . Xu, W. Yu, P. Ghamisi, M. Kopp, and S. Hochreiter, “Txt2img-mhn: Remote sensing image generation from text using modern hopfield networks,”IEEE Transactions on Image Processing, vol. 32, pp. 5737–5750, 2023

    2023

  6. [13]

    Improved techniques for training gans,

    T. Salimans, I. Goodfellow, W. Zaremba, V . Cheung, A. Radford, and X. Chen, “Improved techniques for training gans,”Advances in neural information processing systems, vol. 29, 2016. 11

    2016

  7. [14]

    Diffu- sionsat: A generative foundation model for satellite imagery,

    S. Khanna, P. Liu, L. Zhou, C. Meng, R. Rombach, M. Burke, D. Lobell, and S. Ermon, “Diffu- sionsat: A generative foundation model for satellite imagery,”arXiv preprint arXiv:2312.03606, 2023

    2023 arXiv

  8. [15]

    Crs-diff: Controllable remote sensing image generation with diffusion model,

    D. Tang, X. Cao, X. Hou, Z. Jiang, J. Liu, and D. Meng, “Crs-diff: Controllable remote sensing image generation with diffusion model,”IEEE Transactions on Geoscience and Remote Sensing, 2024

    2024

  9. [16]

    Satsynth: Augmenting image- mask pairs through diffusion models for aerial semantic segmentation,

    A. Toker, M. Eisenberger, D. Cremers, and L. Leal-Taixé, “Satsynth: Augmenting image- mask pairs through diffusion models for aerial semantic segmentation,” inProceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pp. 27695–27705, 2024

    2024

  10. [17]

    Aerogen: enhancing remote sensing object detection with diffusion-driven data generation,

    D. Tang, X. Cao, X. Wu, J. Li, J. Yao, X. Bai, D. Jiang, Y . Li, and D. Meng, “Aerogen: enhancing remote sensing object detection with diffusion-driven data generation,”arXiv preprint arXiv:2411.15497, 2024

    2024 arXiv

  11. [18]

    Mmo-ig: Multi-class and multi-scale object image generation for remote sensing,

    C. Yang, B. Zhao, Q. Zhou, and Q. Wang, “Mmo-ig: Multi-class and multi-scale object image generation for remote sensing,”IEEE Transactions on Geoscience and Remote Sensing, 2025

    2025

  12. [20]

    Diffusion models meet remote sensing: Principles, methods, and perspectives,

    Y . Liu, J. Yue, S. Xia, P. Ghamisi, W. Xie, and L. Fang, “Diffusion models meet remote sensing: Principles, methods, and perspectives,”IEEE Transactions on Geoscience and Remote Sensing, 2024

    2024

  13. [21]

    Dual-diffusion: Dual conditional denoising diffusion probabilistic models for blind super-resolution reconstruction in rsis,

    M. Xu, J. Ma, and Y . Zhu, “Dual-diffusion: Dual conditional denoising diffusion probabilistic models for blind super-resolution reconstruction in rsis,”IEEE Geoscience and Remote Sensing Letters, vol. 20, pp. 1–5, 2023

    2023

  14. [22]

    A multiscale generalized shrinkage threshold network for image blind deblurring in remote sensing,

    Y . Feng, Y . Yang, X. Fan, Z. Zhang, and J. Zhang, “A multiscale generalized shrinkage threshold network for image blind deblurring in remote sensing,”IEEE Transactions on Geoscience and Remote Sensing, vol. 62, pp. 1–16, 2024

    2024

  15. [23]

    Diffusion models for spatio-temporal-spectral fusion of homogeneous gaofen-1 satellite platforms,

    J. Wei, L. Gan, W. Tang, M. Li, and Y . Song, “Diffusion models for spatio-temporal-spectral fusion of homogeneous gaofen-1 satellite platforms,”International Journal of Applied Earth Observation and Geoinformation, vol. 128, p. 103752, 2024

    2024

  16. [24]

    Hyperspectral and panchromatic images fusion based on the dual conditional diffusion models,

    S. Li, S. Li, and L. Zhang, “Hyperspectral and panchromatic images fusion based on the dual conditional diffusion models,”IEEE Transactions on Geoscience and Remote Sensing, vol. 61, pp. 1–15, 2023

    2023

  17. [25]

    Ddpm-cd: Remote sensing change detection using denoising diffusion probabilistic models,

    W. G. C. Bandara, N. G. Nair, and V . M. Patel, “Ddpm-cd: Remote sensing change detection using denoising diffusion probabilistic models,”arXiv preprint arXiv:2206.11892, vol. 3, 2022

    2022 arXiv

  18. [26]

    Pred- iff: Precipitation nowcasting with latent diffusion models,

    Z. Gao, X. Shi, B. Han, H. Wang, X. Jin, D. Maddix, Y . Zhu, M. Li, and Y . B. Wang, “Pred- iff: Precipitation nowcasting with latent diffusion models,”Advances in Neural Information Processing Systems, vol. 36, pp. 78621–78656, 2023

    2023

  19. [27]

    Geosynth: Contextually-aware high-resolution satellite image synthesis,

    S. Sastry, S. Khanal, A. Dhakal, and N. Jacobs, “Geosynth: Contextually-aware high-resolution satellite image synthesis,” inProceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pp. 460–470, 2024

    2024

  20. [28]

    Advancing controllable diffusion model for few-shot object detection in optical remote sensing imagery,

    T. Zhang, Y . Zhuang, X. Zhang, G. Wang, H. Chen, and F. Bi, “Advancing controllable diffusion model for few-shot object detection in optical remote sensing imagery,” inIGARSS 2024-2024 IEEE International Geoscience and Remote Sensing Symposium, pp. 7600–7603, IEEE, 2024

    2024

  21. [29]

    Controllable generative knowledge driven few-shot object detection from optical remote sensing imagery,

    T. Zhang, Y . Zhuang, G. Wang, H. Chen, H. Wang, L. Li, and J. Li, “Controllable generative knowledge driven few-shot object detection from optical remote sensing imagery,”IEEE Transactions on Geoscience and Remote Sensing, 2025

    2025

  22. [30]

    Control copy-paste: Controllable diffusion-based augmentation method for remote sensing few-shot object detection,

    Y . Liu, J. Pan, and B. Zhang, “Control copy-paste: Controllable diffusion-based augmentation method for remote sensing few-shot object detection,”arXiv preprint arXiv:2507.21816, 2025. 12

    2025 arXiv

  23. [31]

    Domain-rag: Retrieval-guided compositional image generation for cross-domain few- shot object detection,

    Y . Li, X. Qiu, Y . Fu, J. Chen, T. Qian, X. Zheng, D. P. Paudel, Y . Fu, X. Huang, L. Van Gool, et al., “Domain-rag: Retrieval-guided compositional image generation for cross-domain few- shot object detection,”arXiv preprint arXiv:2506.05872, 2025

    2025

  24. [32]

    Styleadv: Meta style adversarial training for cross-domain few-shot learning,

    Y . Fu, Y . Xie, Y . Fu, and Y .-G. Jiang, “Styleadv: Meta style adversarial training for cross-domain few-shot learning,” inProceedings of the IEEE/CVF conference on computer vision and pattern recognition, pp. 24575–24584, 2023

    2023

  25. [33]

    Ntire 2025 challenge on cross-domain few-shot object detection: methods and results,

    Y . Fu, X. Qiu, B. Ren, Y . Fu, R. Timofte, N. Sebe, M.-H. Yang, L. Van Gool,et al., “Ntire 2025 challenge on cross-domain few-shot object detection: methods and results,” inCVPRW, 2025

    2025

  26. [34]

    Out-of-domain robustness via targeted augmentations,

    I. Gao, S. Sagawa, P. W. Koh, T. Hashimoto, and P. Liang, “Out-of-domain robustness via targeted augmentations,” inInternational Conference on Machine Learning, pp. 10800–10834, PMLR, 2023

    2023

  27. [35]

    Reducing semantic confusion: Scene-aware aggregation network for remote sensing cross-modal retrieval,

    J. Pan, Q. Ma, and C. Bai, “Reducing semantic confusion: Scene-aware aggregation network for remote sensing cross-modal retrieval,” inProceedings of the 2023 ACM International Conference on Multimedia Retrieval, pp. 398–406, 2023

    2023

  28. [36]

    A prior instruction representation framework for remote sensing image-text retrieval,

    J. Pan, Q. Ma, and C. Bai, “A prior instruction representation framework for remote sensing image-text retrieval,” inProceedings of the 31st ACM International Conference on Multimedia, pp. 611–620, 2023

    2023

  29. [37]

    Pir: Remote sensing image-text retrieval with prior instruction representation learning,

    J. Pan, M. Ma, Q. Ma, C. Bai, and S. Chen, “Pir: Remote sensing image-text retrieval with prior instruction representation learning,”arXiv preprint arXiv:2405.10160, 2024

    2024 arXiv

  30. [38]

    Simple copy-paste is a strong data augmentation method for instance segmentation,

    G. Ghiasi, Y . Cui, A. Srinivas, R. Qian, T.-Y . Lin, E. D. Cubuk, Q. V . Le, and B. Zoph, “Simple copy-paste is a strong data augmentation method for instance segmentation,” inProceedings of the IEEE/CVF conference on computer vision and pattern recognition, pp. 2918–2928, 2021

    2021

  31. [39]

    Generalized odin: Detecting out-of-distribution image without learning from out-of-distribution data,

    Y .-C. Hsu, Y . Shen, H. Jin, and Z. Kira, “Generalized odin: Detecting out-of-distribution image without learning from out-of-distribution data,” inProceedings of the IEEE/CVF conference on computer vision and pattern recognition, pp. 10951–10960, 2020

    2020

  32. [40]

    Direction-oriented visual–semantic embedding model for remote sensing image–text retrieval,

    Q. Ma, J. Pan, and C. Bai, “Direction-oriented visual–semantic embedding model for remote sensing image–text retrieval,”IEEE Transactions on Geoscience and Remote Sensing, vol. 62, pp. 1–14, 2024

    2024

  33. [41]

    On the importance of gradients for detecting distributional shifts in the wild,

    R. Huang, A. Geng, and Y . Li, “On the importance of gradients for detecting distributional shifts in the wild,”Advances in Neural Information Processing Systems, 2021

    2021

  34. [42]

    U-net: Convolutional networks for biomedical image segmentation,

    O. Ronneberger, P. Fischer, and T. Brox, “U-net: Convolutional networks for biomedical image segmentation,” inMedical Image Computing and Computer-Assisted Intervention–MICCAI 2015: 18th International Conference, Munich, Germany, October 5-9, 2015, Proceedings, Part III 18, pp...

    2015

  35. [43]

    Adding conditional control to text-to-image diffu- sion models,

    L. Zhang, A. Rao, and M. Agrawala, “Adding conditional control to text-to-image diffu- sion models,” inProceedings of the IEEE/CVF international conference on computer vision, pp. 3836–3847, 2023

    2023

  36. [44]

    Instancediffusion: Instance- level control for image generation,

    X. Wang, T. Darrell, S. S. Rambhatla, R. Girdhar, and I. Misra, “Instancediffusion: Instance- level control for image generation,” inProceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pp. 6232–6242, 2024

    2024

  37. [45]

    Automatic segmentation and fitting of image edge contours based on douglas algorithm,

    Y . Cao and M. Wang, “Automatic segmentation and fitting of image edge contours based on douglas algorithm,” in2022 IEEE Conference on Telecommunications, Optics and Computer Science (TOCS), pp. 451–455, IEEE, 2022

    2022

  38. [46]

    Openearthmap: A benchmark dataset for global high-resolution land cover mapping,

    J. Xia, N. Yokoya, B. Adriano, and C. Broni-Bediako, “Openearthmap: A benchmark dataset for global high-resolution land cover mapping,” inProceedings of the IEEE/CVF Winter Conference on Applications of Computer Vision, pp. 6254–6264, 2023

    2023

  39. [47]

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

    J. Wang, Z. Zheng, A. Ma, X. Lu, and Y . Zhong, “Loveda: A remote sensing land-cover dataset for domain adaptive semantic segmentation,”arXiv preprint arXiv:2110.08733, 2021. 13

    2021 arXiv

  40. [48]

    Deepglobe 2018: A challenge to parse the earth through satellite images,

    I. Demir, K. Koperski, D. Lindenbaum, G. Pang, J. Huang, S. Basu, F. Hughes, D. Tuia, and R. Raskar, “Deepglobe 2018: A challenge to parse the earth through satellite images,” in Proceedings of the IEEE conference on computer vision and pattern recognition workshops, pp. 172–181, 2018

    2018

  41. [49]

    Samrs: Scaling-up remote sensing segmentation dataset with segment anything model,

    D. Wang, J. Zhang, B. Du, M. Xu, L. Liu, D. Tao, and L. Zhang, “Samrs: Scaling-up remote sensing segmentation dataset with segment anything model,”Advances in Neural Information Processing Systems, vol. 36, pp. 8815–8827, 2023

    2023

  42. [50]

    Locate anything on earth: Advancing open-vocabulary object detection for remote sensing community,

    J. Pan, Y . Liu, Y . Fu, M. Ma, J. Li, D. P. Paudel, L. Van Gool, and X. Huang, “Locate anything on earth: Advancing open-vocabulary object detection for remote sensing community,” in Proceedings of the AAAI Conference on Artificial Intelligence, vol. 39, pp. 6281–6289, 2025

    2025

  43. [51]

    Remote sensing image scene classification: Benchmark and state of the art,

    G. Cheng, J. Han, and X. Lu, “Remote sensing image scene classification: Benchmark and state of the art,”Proceedings of the IEEE, vol. 105, no. 10, pp. 1865–1883, 2017

    2017

  44. [52]

    Object detection in 20 years: A survey,

    Z. Zou, K. Chen, Z. Shi, Y . Guo, and J. Ye, “Object detection in 20 years: A survey,”Proceedings of the IEEE, vol. 111, no. 3, pp. 257–276, 2023

    2023

  45. [53]

    A review of semantic segmentation using deep neural networks,

    Y . Guo, Y . Liu, T. Georgiou, and M. S. Lew, “A review of semantic segmentation using deep neural networks,”International journal of multimedia information retrieval, vol. 7, pp. 87–93, 2018

    2018

  46. [54]

    Learning transferable visual models from natural language supervi- sion,

    A. Radford, J. W. Kim, C. Hallacy, A. Ramesh, G. Goh, S. Agarwal, G. Sastry, A. Askell, P. Mishkin, J. Clark,et al., “Learning transferable visual models from natural language supervi- sion,” inInternational conference on machine learning, pp. 8748–8763, PmLR, 2021

    2021

  47. [55]

    Detecting twenty-thousand classes using image-level supervision,

    X. Zhou, R. Girdhar, A. Joulin, P. Krähenbühl, and I. Misra, “Detecting twenty-thousand classes using image-level supervision,” inEuropean conference on computer vision, pp. 350–368, Springer, 2022

    2022

  48. [56]

    Detect everything with few examples,

    X. Zhang, Y . Liu, Y . Wang, and A. Boularias, “Detect everything with few examples,”arXiv preprint arXiv:2309.12969, 2023

    2023 arXiv

  49. [57]

    Cross-domain few-shot object detection via enhanced open-set object detector,

    Y . Fu, Y . Wang, Y . Pan, L. Huai, X. Qiu, Z. Shangguan, T. Liu, Y . Fu, L. Van Gool, and X. Jiang, “Cross-domain few-shot object detection via enhanced open-set object detector,” inEuropean Conference on Computer Vision, pp. 247–264, Springer, 2024

    2024

  50. [58]

    Grounding dino: Marrying dino with grounded pre-training for open-set object detection,

    S. Liu, Z. Zeng, T. Ren, F. Li, H. Zhang, J. Yang, Q. Jiang, C. Li, J. Yang, H. Su,et al., “Grounding dino: Marrying dino with grounded pre-training for open-set object detection,” in European Conference on Computer Vision, pp. 38–55, Springer, 2024

    2024

  51. [59]

    Enhance then search: An augmentation-search strategy with foundation models for cross-domain few-shot object detec- tion,

    J. Pan, Y . Liu, X. He, L. Peng, J. Li, Y . Sun, and X. Huang, “Enhance then search: An augmentation-search strategy with foundation models for cross-domain few-shot object detec- tion,” 2025

    2025

  52. [60]

    Exploring models and data for remote sensing image caption generation,

    X. Lu, B. Wang, X. Zheng, and X. Li, “Exploring models and data for remote sensing image caption generation,”IEEE Transactions on Geoscience and Remote Sensing, vol. 56, no. 4, pp. 2183–2195, 2017

    2017

  53. [61]

    Towards open-vocabulary remote sensing image semantic segmentation,

    C. Ye, Y . Zhuge, and P. Zhang, “Towards open-vocabulary remote sensing image semantic segmentation,” inProceedings of the AAAI Conference on Artificial Intelligence, vol. 39, pp. 9436–9444, 2025

    2025

  54. [62]

    Object detection in optical remote sensing images: A survey and a new benchmark,

    K. Li, G. Wan, G. Cheng, L. Meng, and J. Han, “Object detection in optical remote sensing images: A survey and a new benchmark,”ISPRS journal of photogrammetry and remote sensing, vol. 159, pp. 296–307, 2020

    2020

  55. [63]

    Dota: A large-scale dataset for object detection in aerial images,

    G.-S. Xia, X. Bai, J. Ding, Z. Zhu, S. Belongie, J. Luo, M. Datcu, M. Pelillo, and L. Zhang, “Dota: A large-scale dataset for object detection in aerial images,” inProceedings of the IEEE conference on computer vision and pattern recognition, pp. 3974–3983, 2018

    2018

  56. [64]

    2d semantic labeling potsdam dataset

    ISPRS, “2d semantic labeling potsdam dataset.” https://www.isprs.org/education/ benchmarks/UrbanSemLab/2d-sem-label-potsdam.aspx, 2013. Accessed: 2024-08-11. 14

    2013

  57. [65]

    Floodnet: A high resolution aerial imagery dataset for post flood scene understanding,

    M. Rahnemoonfar, T. Chowdhury, A. Sarkar, D. Varshney, M. Yari, and R. R. Murphy, “Floodnet: A high resolution aerial imagery dataset for post flood scene understanding,”IEEE Access, vol. 9, pp. 89644–89654, 2021

    2021

  58. [66]

    Flair: a country-scale land cover semantic segmentation dataset from multi-source optical imagery,

    A. Garioud, N. Gonthier, L. Landrieu, A. De Wit, M. Valette, M. Poupée, S. Giordano,et al., “Flair: a country-scale land cover semantic segmentation dataset from multi-source optical imagery,”Advances in Neural Information Processing Systems, vol. 36, pp. 16456–16482, 2023

    2023

  59. [67]

    Gans trained by a two time-scale update rule converge to a local nash equilibrium,

    M. Heusel, H. Ramsauer, T. Unterthiner, B. Nessler, and S. Hochreiter, “Gans trained by a two time-scale update rule converge to a local nash equilibrium,”Advances in neural information processing systems, vol. 30, 2017

    2017

  60. [68]

    Visualizing data using t-sne.,

    L. Van der Maaten and G. Hinton, “Visualizing data using t-sne.,”Journal of machine learning research, vol. 9, no. 11, 2008

    2008

  61. [69]

    Segment anything,

    A. Kirillov, E. Mintun, N. Ravi, H. Mao, C. Rolland, L. Gustafson, T. Xiao, S. Whitehead, A. C. Berg, W.-Y . Lo,et al., “Segment anything,” inProceedings of the IEEE/CVF international conference on computer vision, pp. 4015–4026, 2023. 15 A Technical Appendices and Supplementa...

    2023

  62. [70]

    Visual Quality ( integer score 0 -100)

  63. [71]

    #$!: ℒ%!

    Semantic Richness ( integer score 0 -100) Image d e s c r i p t i o n : < image d e s c r i p t i o n text > Respond ONLY with exactly the fo ll owi ng format and nothing else : Visual Quality Score : < integer 0 -100 > Semantic Richness Score : < integer 0 -100 > Reason : < o...

Pith tools

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