Pith. sign in

REVIEW 3 major objections 5 minor 1 cited by

SIDA: Synthetic Image Driven Zero-shot Domain Adaptation

T0 review · 3 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 SIDA claims that a small set of synthetic images, generated by prompting a latent diffusion model with vision-language scene descriptions and translated to the target style, can replace text-driven alignment in zero-shot domain…

desk verdict Sensible empirical paper with a real efficiency win, but the headline margins are partly selected on the target validation sets it evaluates on; worth refereeing, but the zero-shot claim needs a fix. read the letter →

arxiv 2507.18632 v2 pith:KGIKLOE6 submitted 2025-07-24 cs.CV cs.AIcs.LGcs.MM

Ye-Chan Kim , SeungJu Cha , Si-Woo Kim , Taewhan Kim , Dong-Jin Kim This is my paper · ORCID
classification cs.CVcs.AIcs.LGcs.MM
keywords zero-shotdomainadaptationsyntheticimagesmixpatchstyletransfersemanticsegmentationdiffusionmodelfeatureadverseweatherdomains
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The reading

The paper tries to show that zero-shot domain adaptation need not rely on text captions like 'Driving at {domain}' to simulate a target domain's style. Instead, a small set of synthetic images is generated from scene descriptions extracted from source images, translated into the target domain, and then used as a statistical proxy for the target's appearance. The central claim is that two feature-level operations built on these synthetic images—Domain Mix, which blends channel-wise style statistics across domains, and Patch Style Transfer, which applies different styles to different patches—let classifier-only fine-tuning outperform text-driven zero-shot adaptation. If correct, this makes adaptation to rare or dangerous domains (fire, sandstorm) both more accurate and far cheaper, because the costly per-image text-alignment loop is removed.

What carries the argument

The load-bearing mechanism is the low-level feature map from the first layer of the segmentation backbone, whose channel-wise mean and standard deviation encode style. Domain Mix takes those statistics from a main translated synthetic image and an auxiliary domain image, blends them with per-channel random weights, and adds Gaussian noise to produce many global style intensities from a few images. Patch Style Transfer splits the source feature into non-overlapping patches and applies AdaIN to each patch using a different Domain-Mix-derived style, so local regions of one image receive different intensities. The final weighted cross-entropy loss up-weights high-entropy stylized samples, pushing the classifier to learn the novel style distribution.

What would settle it

Run SIDA and a text-driven baseline on a target domain whose visual style the fixed 'Driving at {domain}' translation prompt cannot reproduce—for example, a domain defined by a sensor artifact or an unseen precipitation type—and compare mIoU; if the gap collapses or reverses, the synthetic proxy is not actually supplying the target style. A second check: replace the target-translated images with randomly prompted synthetic images; if accuracy is unchanged, Domain Mix and Patch Style Transfer are acting as generic regularization, not as target-style transfer.

Watch

Extended reading notes

Core claim

The paper's central discovery is that the first-layer channel-wise mean and standard deviation of a few diffusion-translated synthetic images, once diversified, are a better proxy for an unseen target domain than the text-aligned style vectors used by prior zero-shot methods. Concretely, with only three synthetic images per domain, SIDA raises mean mIoU on Cityscapes-to-ACDC from ULDA's 42.47 to 43.13, and on Cityscapes-to-Fire/Sandstorm from 22.62 to 25.08. The same pipeline cuts adaptation time disproportionately: the time ratio when moving from the small Cityscapes source to the large GTA5 source is 1.47 for SIDA versus 25.30 for the full ULDA procedure, because no per-image optimization is needed. The method's two modules—Domain Mix and Patch Style Transfer—are what turn a small synthetic set into a wide range of global and local style intensities rather than a single fixed style.

Load-bearing premise

The whole method rests on the assumption that the channel-wise mean and standard deviation of a few diffusion-translated synthetic images, after mixing with an auxiliary domain and adding noise, stand in for the real target domain's feature statistics closely enough that fine-tuning the classifier on them transfers.

Editorial extensions

If this is right

  • Classifier-only fine-tuning on stylized features is enough: the text-image alignment loop of prior methods disappears, so adaptation cost stops growing with source dataset size.
  • A handful of synthetic images (even one) can serve as a target proxy; the number of generated images is not the main driver of performance.
  • Domain Mix and Patch Style Transfer transfer to text-driven baselines, improving PODA's and ULDA's accuracy when added on top, so they are general feature-augmentation tools rather than tied to synthetic generation.
  • The method is strongest where text descriptions fail most—fire and sandstorm—so image-driven proxies are the more promising route for rare-domain adaptation.

Reading between the lines

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

  • The ceiling of methods like SIDA is set by the generative model's ability to render a target domain; as image generation improves, zero-shot adaptation could improve without any new adaptation machinery.
  • The same style-statistic augmentation recipe could extend to other dense prediction tasks (object detection, depth estimation) and to non-driving scenes, since the pipeline does not assume a particular semantic class set.
  • Because a single synthetic image already beats the text baseline, the bottleneck appears to be style-statistic diversity, not sample count; an explicit test would be to maximize diversity of generated styles rather than number of images.
  • If the synthetic proxy is accurate enough, ZSDA could be applied to genuinely rare domains where collecting real images is unsafe, by generating plausible worst-case appearances from textual descriptions of hazards.
Share X Bluesky LinkedIn Reddit HN

Signed reviews

No signed human review yet.

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

3 major / 5 minor

Summary. This paper addresses zero-shot domain adaptation for semantic segmentation, where no target-domain images are available at adaptation time. The proposed SIDA method replaces the text-driven learned style vectors of PØDA/ULDA with a small set of synthetic images: a VLM (GPT-4o) generates scene descriptions from source images, Stable Diffusion generates source-like images, and image-to-image translation with prompts such as 'Driving at {domain}' produces translated images for each target domain. Channel-wise mean/standard-deviation statistics of these translated images are blended with an auxiliary domain (Domain Mix, Eq. 3) and Gaussian noise (Eq. 4), then applied patch-wise to source features via AdaIN (Eqs. 6-8); the classifier is fine-tuned with an entropy-weighted cross-entropy loss (Eqs. 9-10). Experiments on Cityscapes→ACDC and Cityscapes→Fire/Sandstorm report gains over ULDA, e.g., 43.13 vs 42.47 mean mIoU on ACDC and 25.08 vs 22.62 on Fire/Sandstorm, together with lower adaptation time scaling (T_GTA/T_CS = 1.47).

Significance. If the results survive a clean evaluation protocol, the paper makes a useful contribution: it shows that a handful of off-the-shelf generative-model images can serve as a target-style proxy, avoiding per-image text alignment and reducing the growth of adaptation time with source size. The method is simple, the component ablations are informative, and the five-seed runs for the main table are good practice. However, the headline numerical advantage over ULDA is currently not established because key hyperparameters are selected on the same ACDC validation set used for reporting; the advantage is comparable in size to the variation across the N sweep in Table 4. The efficiency analysis is also suggestive rather than conclusive because the re-implemented ULDA* is not assigned a reported mIoU.

major comments (3)
  1. [Section 4.1, Fig. 5, Table 4] The zero-shot claim is compromised by target-validation-based hyperparameter selection. Section 4.1 fixes N=3, m=3, s_e^2=0.075^2, and tau_ent=1.0; Fig. 5 selects s_e by maximizing CS→ACDC mean mIoU, and Table 4 selects N by the same criterion on the same ACDC validation set that is reported in Table 1. This is a selection leak: Table 4 shows N=10 yields 42.44, below ULDA's 42.47, and N=100 yields 42.53, so the selected N=3 is the only configuration above 42.6. The 0.66-point headline margin in Table 1 is therefore within the range that hyperparameter selection can produce. The authors should either use a held-out split of ACDC for tuning with a separate test split for reporting, or pre-register target-free selection rules (e.g., source-only or auxiliary-domain criteria), and report the full N and s_e sweeps under the corrected protocol.
  2. [Section 4.4, Fig. 7] The feature-visualization evidence in Fig. 7 cannot independently validate the target proxy: the t-SNE is computed on ACDC style features from the same validation split used to tune s_e and N, and the evaluation is qualitative. Please provide a quantitative distributional metric (e.g., MMD or Fréchet distance on style statistics) computed on a target split not used for any selection, and include per-domain numbers rather than a single t-SNE plot.
  3. [Section 4.4, Table 6] The claim that the proposed components consistently improve existing methods is not fully supported: ULDA+SIDA decreases Rain by 0.07 mIoU (35.84→35.77). The text says 'generally,' which is acceptable, but the abstract and conclusion should temper unconditional statements such as 'consistently outperforms,' or the authors should analyze why the component transfer hurts on Rain.
minor comments (5)
  1. [Section 4.4] In the last paragraph, 'this stems from theSIDA' contains a missing space; please fix the typo.
  2. [Table 1] The row labels 'GTA5' and 'GTA5 Cityscapes' are ambiguous; clarify which entry is the source and which is the target for each block.
  3. [Figure 7] The legend uses three bullet symbols that are not visually distinct in the caption; please replace them with clearly distinct markers and a more descriptive caption.
  4. [Eq. (3)] The term 1_c - lambda should be explicitly defined as an elementwise complement; currently 1_c is only described as a vector of all ones.
  5. [Section 4.3] The phrase 'the noise range of s_e(0.075~0.1)' should be written as 's_e in [0.075, 0.1]' for precision.

Circularity Check

2 steps flagged · score 6.0 of 10

The headline SOTA margin is partly constructed by selecting hyperparameters (N, noise) on the same ACDC validation set used for evaluation; no equation-level circularity found.

  1. fitted input called prediction [Section 4.3, 'The number of synthetic images' and Table 4]
    "Table 4 illustrates the results of the ablation analysis on the number of synthetic images used during adaptation. In most cases, even with an extremely limited number of images, our method consistently outperforms the state-of-the-art ULDA [31]. In our experiments, we utilized three images that yielded the highest performance."

    The number of synthetic images N is selected by maximizing the very CS-to-ACDC mean-mIoU that is then reported as SIDA's headline advantage (43.13) over ULDA (42.47). Table 4 shows that N=10 yields 42.44, which is below ULDA, and N=100 yields only 42.53; N=3 is the peak of the same tuning curve used as the test. Thus the claimed improvement over ULDA exists only at the selected point, making the reported superiority partly an artifact of choosing the evaluation-set maximum. Since the paper defines ZSDA as adaptation without target domain image data, using the target validation set to select N and then reporting its score is a form of fitted input being presented as a zero-shot result.

  2. fitted input called prediction [Section 4.3, 'Impact of noise variance' and Fig. 5]
    "Fig. 5 shows the hyperparameter analysis on the Gaussian noise 𝝐, 𝝐′ with varying variance, added to Domain Mix and Patch Style Transfer. We use 𝑠𝑒 = 0.075, which shows the best performance."

    The noise scale s_e is chosen as the maximizer of the same ACDC mean-mIoU curve that is later used to claim state-of-the-art performance, with Fig. 5 plotting 'Mean mIoU' against the ULDA mean line. The subsequent statement that SIDA outperforms ULDA is conditional on a hyperparameter fitted to the evaluation set, so the margin is partly constructed by selection. This is not an equation-level circularity, but it is evaluation-set fitting presented as a zero-shot prediction.

full rationale

No equation-level circularity is present: the method never defines its target in terms of its own output, no result is derived by fitting a parameter into a renamed prediction, and the self-citations in the related work are not load-bearing. The synthetic-image proxy, Domain Mix, Patch Style Transfer, and entropy-weighted loss are independent algorithmic components with their own ablation support. However, the central empirical claim of consistent SOTA performance is partially circular in the selection sense: Fig. 5 and Table 4 choose the noise level and the number of synthetic images by maximizing the same CS-to-ACDC mean-mIoU used in the main comparison, and Table 4 itself shows that other choices (N=10) fall below ULDA. The paper also evaluates on 'the validation sets' of target domains, which are the same sets used for this hyperparameter selection. Therefore the 0.66-point margin over ULDA is not a clean out-of-sample prediction; it is the selected maximum of a tuning curve. I score this as partial circularity (6) rather than full circularity because the core method has independent content and the reduction is empirical selection, not an equation-level identity.

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

All central ingredients are empirical modeling choices. The method borrows AdaIN and style-statistics mixing from prior literature, and the main unverified premises are the fidelity of the synthetic style proxy and the usefulness of entropy weighting. No new physical entities are introduced. Hyperparameters N, s_e, tau_ent, and patch size are hand-selected, and N and s_e are tuned on the target validation set, so the ledger entries do not provide independent evidence.

free parameters (5)
  • Number of synthetic images per domain N = 3
    Selected from Table 4 because N=3 gives the highest CS to ACDC mean mIoU (43.13); values 1, 5, 10, and 100 give 42.53, 42.51, 42.44, and 42.53, so N=3 is a benchmark-tuned choice.
  • Noise variance s_e squared = 0.075 squared
    Fig. 5 sweeps the noise level and selects 0.075 as best on the same ACDC validation set used for final numbers; the paper notes that values between 0.075 and 0.1 beat ULDA.
  • Patch grid size m = 3 (m by m)
    Set in Sec. 4.1; no ablation study is reported for m, so its value is hand-chosen and may affect local-style diversity.
  • Entropy threshold tau_ent = 1.0
    Set in Sec. 4.1; no ablation is reported, and the entropy-weighting hypothesis is only indirectly validated through the full-pipeline ablation.
  • Mixing weight lambda distribution = uniform in [0, 1] per channel
    Domain Mix samples per-channel blending weights from this hand-chosen distribution; the distribution shape controls how much auxiliary style enters the proxy but is not optimized or ablated.
assumptions (5)
  • domain assumption First-layer channel-wise mean and standard deviation are sufficient domain-style descriptors, and AdaIN can transfer them to source features.
    Inherited from PØDA, ULDA, and AdaIN without proof; Eqs. (5)-(8) apply this assumption directly.
  • domain assumption Translated synthetic images generated by Stable Diffusion v1.5 preserve source semantics while carrying target style.
    Sec. 3.1 assumes the image-to-image step with the prompt 'Driving at {domain}' yields images whose low-level statistics are target-like; Fig. 6 shows examples but no quantitative semantic-consistency check.
  • domain assumption Blending main and auxiliary domain style statistics plus Gaussian noise spans the real target style distribution.
    Core of Domain Mix in Eqs. (3)-(4); tested only via downstream mIoU and qualitative t-SNE, not by a distributional metric.
  • domain assumption High-entropy predictions from the source classifier indicate uncertain target-style samples worth upweighting.
    Sec. 3.3 hypothesizes this; no independent experiment isolates the entropy mechanism from the rest of the pipeline.
  • domain assumption Auxiliary domain selection by cosine similarity in feature space identifies a useful blending partner for Domain Mix.
    Eq. (1) chooses the domain with highest cosine similarity to the main domain feature; no ablation tests whether this selection rule is necessary or whether random auxiliary choice would work.

how reviews work

0 comments
Cite this review

Pith. "Pith review of SIDA: Synthetic Image Driven Zero-shot Domain Adaptation." pith.science (2026). https://pith.science/paper/KGIKLOE6

@misc{pith2026250718632,
  author       = {Pith},
  title        = {Pith review of: SIDA: Synthetic Image Driven Zero-shot Domain Adaptation},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/KGIKLOE6}},
  note         = {Machine review of arXiv:2507.18632}
}
read the original abstract

Zero-shot domain adaptation is a method for adapting a model to a target domain without utilizing target domain image data. To enable adaptation without target images, existing studies utilize CLIP's embedding space and text description to simulate target-like style features. Despite the previous achievements in zero-shot domain adaptation, we observe that these text-driven methods struggle to capture complex real-world variations and significantly increase adaptation time due to their alignment process. Instead of relying on text descriptions, we explore solutions leveraging image data, which provides diverse and more fine-grained style cues. In this work, we propose SIDA, a novel and efficient zero-shot domain adaptation method leveraging synthetic images. To generate synthetic images, we first create detailed, source-like images and apply image translation to reflect the style of the target domain. We then utilize the style features of these synthetic images as a proxy for the target domain. Based on these features, we introduce Domain Mix and Patch Style Transfer modules, which enable effective modeling of real-world variations. In particular, Domain Mix blends multiple styles to expand the intra-domain representations, and Patch Style Transfer assigns different styles to individual patches. We demonstrate the effectiveness of our method by showing state-of-the-art performance in diverse zero-shot adaptation scenarios, particularly in challenging domains. Moreover, our approach achieves high efficiency by significantly reducing the overall adaptation time.

Figures

Figures reproduced from arXiv: 2507.18632 by the authors.

Figure 1
Figure 1. Global and local style intensity variations in real [PITH_FULL_IMAGE:figures/full_fig_p001_1.png] view at source ↗
Figure 2
Figure 2. Pipeline of SIDA. Our method applies Domain Mix to the synthetic image features generated for each domain, producing diverse target-like styles. Patch Style Transfer then leverages these features to apply locally varying styles to each patch, enabling fine-grained style diversity. In this section, we describe how SIDA generates and utilizes synthetic images in ZSDA. Rather than relying on fixed text de￾scriptions, o… view at source ↗
Figure 3
Figure 3. Illustration of image generation process. In the [PITH_FULL_IMAGE:figures/full_fig_p003_3.png] view at source ↗
Figures from the paper (4 more)
Figure 4
Figure 4. Figure 4: Qualitative results on Fire and Sandstorm domains. (a) and (b) show qualitative results obtained on the Sand-Fire [PITH_FULL_IMAGE:figures/full_fig_p006_4.png]
Figure 5
Figure 5. Figure 5: Ablation study on noise hyperparameters of [PITH_FULL_IMAGE:figures/full_fig_p006_5.png]
Figure 6
Figure 6. Figure 6: Qualitative results for each generation method across four target domains—Night, Snow, Rain, and Fog (from top [PITH_FULL_IMAGE:figures/full_fig_p007_6.png]
Figure 7
Figure 7. Figure 7: The visualization results of the simulated target [PITH_FULL_IMAGE:figures/full_fig_p008_7.png]

Review history (2 revisions) →

Discussion (0). Continue with ORCID to comment.

Forward citations

Cited by 1 Pith paper

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

  1. Sali4Vid: Saliency-Aware Video Reweighting and Adaptive Caption Retrieval for Dense Video Captioning

    cs.CV 2025-09 conditional novelty 6.0 of 10

    Sali4Vid improves dense video captioning by reweighting video features with timestamp-derived sigmoid importance and adaptively retrieving captions per semantic segment, achieving new SOTA on YouCook2 and ViTT.

Reference graph

Works this paper leans on

39 extracted references · 23 canonical work pages · cited by 1 Pith paper

  1. [1]

    Josh Achiam, Steven Adler, Sandhini Agarwal, Lama Ahmad, Ilge Akkaya, Floren- cia Leoni Aleman, Diogo Almeida, Janko Altenschmidt, Sam Altman, Shyamal Anadkat, et al. 2023. Gpt-4 technical report. arXiv preprint arXiv:2303.08774 (2023)

    arXiv 2023

  2. [2]

    Yasser Benigmim, Subhankar Roy, Slim Essid, Vicky Kalogeiton, and Stéphane Lathuilière. 2023. One-shot unsupervised domain adaptation with personalized diffusion models. In Proceedings of the IEEE/CVF conference on computer vision and pattern recognition. 698–708

    work page 2023

  3. [3]

    SeungJu Cha, Kwanyoung Lee, Ye-Chan Kim, Hyunwoo Oh, and Dong-Jin Kim

  4. [4]

    Liang-Chieh Chen, Yukun Zhu, George Papandreou, Florian Schroff, and Hartwig Adam. 2018. Encoder-decoder with atrous separable convolution for semantic image segmentation. In Proceedings of the European conference on computer vision (ECCV). 801–818

    2018

  5. [5]

    Marius Cordts, Mohamed Omran, Sebastian Ramos, Timo Rehfeld, Markus En- zweiler, Rodrigo Benenson, Uwe Franke, Stefan Roth, and Bernt Schiele. 2016. The cityscapes dataset for semantic urban scene understanding. In Proceedings of the IEEE conference on computer vision and pattern recognition . 3213–3223

    2016

  6. [6]

    Mohammad Fahes, Tuan-Hung Vu, Andrei Bursuc, Patrick Pérez, and Raoul De Charette. 2023. Poda: Prompt-driven zero-shot domain adaptation. InProceed- ings of the IEEE/CVF International Conference on Computer Vision . 18623–18633

    work page 2023

  7. [7]

    Qi Fan, Mattia Segu, Yu-Wing Tai, Fisher Yu, Chi-Keung Tang, Bernt Schiele, and Dengxin Dai. 2023. Towards robust object detection invariant to real-world do- main shifts. In The Eleventh International Conference on Learning Representations (ICLR 2023). OpenReview

    work page 2023

  8. [8]

    Yaroslav Ganin, Evgeniya Ustinova, Hana Ajakan, Pascal Germain, Hugo Larochelle, François Laviolette, Mario March, and Victor Lempitsky. 2016. Domain-adversarial training of neural networks. Journal of machine learning research 17, 59 (2016), 1–35

    2016

Show all 39 references
  1. [9]

    Yaru Hao, Zewen Chi, Li Dong, and Furu Wei. 2024. Optimizing prompts for text-to-image generation. Advances in Neural Information Processing Systems 36 (2024)

    2024

  2. [10]

    Judy Hoffman, Eric Tzeng, Taesung Park, Jun-Yan Zhu, Phillip Isola, Kate Saenko, Alexei Efros, and Trevor Darrell. 2018. Cycada: Cycle-consistent adversarial domain adaptation. In International conference on machine learning . Pmlr, 1989– 1998

    2018

  3. [11]

    Lukas Hoyer, Dengxin Dai, and Luc Van Gool. 2022. Daformer: Improving network architectures and training strategies for domain-adaptive semantic segmentation. In Proceedings of the IEEE/CVF conference on computer vision and pattern recognition. 9924–9935

    2022

  4. [12]

    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

  5. [13]

    Si-Woo Kim, MinJu Jeon, Ye-Chan Kim, Soeun Lee, Taewhan Kim, and Dong-Jin Kim. 2025. SynC: Synthetic Image Caption Dataset Refinement with One-to- many Mapping for Zero-shot Image Captioning. In ACM Multimedia 2025

    2025

  6. [14]

    Taewhan Kim, Soeun Lee, Si-Woo Kim, and Dong-Jin Kim. 2025. Vipcap: Retrieval text-based visual prompts for lightweight image captioning. In Proceedings of the AAAI Conference on Artificial Intelligence , Vol. 39. 4320–4328

    2025

  7. [15]

    Gihyun Kwon and Jong Chul Ye. 2022. Clipstyler: Image style transfer with a single text condition. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition . 18062–18071

    2022

  8. [16]

    Soeun Lee, Si-Woo Kim, Taewhan Kim, and Dong-Jin Kim. 2024. Ifcap: Image-like retrieval and frequency-based entity filtering for zero-shot captioning. arXiv preprint arXiv:2409.18046 (2024)

    2024 arXiv

  9. [17]

    Joshua Niemeijer, Manuel Schwonberg, Jan-Aike Termöhlen, Nico M Schmidt, and Tim Fingscheidt. 2024. Generalization by adaptation: Diffusion-based domain extension for domain-generalized semantic segmentation. In Proceedings of the IEEE/CVF Winter Conference on Applications of ...

    2024

  10. [18]

    Hyunwoo Oh, SeungJu Cha, Kwanyoung Lee, Si-Woo Kim, and Dong-Jin Kim

  11. [19]

    Chenghao Qian, Yuhu Guo, Yuhong Mo, and Wenjing Li. 2024. WeatherDG: LLM-assisted Procedural Weather Generation for Domain-Generalized Semantic Segmentation. arXiv preprint arXiv:2410.12075 (2024)

    2024 arXiv

  12. [20]

    In ACM Multimedia 2025

    CatchPhrase: EXPrompt-Guided Encoder Adaptation for Audio-to-Image Generation. In ACM Multimedia 2025

    2025

  13. [21]

    Alec Radford, Jong Wook Kim, Chris Hallacy, Aditya Ramesh, Gabriel Goh, Sand- hini Agarwal, Girish Sastry, Amanda Askell, Pamela Mishkin, Jack Clark, et al

  14. [22]

    Fengchun Qiao, Long Zhao, and Xi Peng. 2020. Learning to learn single domain generalization. In Proceedings of the IEEE/CVF conference on computer vision and pattern recognition. 12556–12565

    2020

  15. [23]

    Stephan R Richter, Vibhav Vineet, Stefan Roth, and Vladlen Koltun. 2016. Playing for data: Ground truth from computer games. In Computer Vision–ECCV 2016: 14th European Conference, Amsterdam, The Netherlands, October 11-14, 2016, Proceedings, Part II 14 . Springer, 102–118

    2016

  16. [24]

    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

  17. [25]

    Aditya Ramesh, Mikhail Pavlov, Gabriel Goh, Scott Gray, Chelsea Voss, Alec Rad- ford, Mark Chen, and Ilya Sutskever. 2021. Zero-shot text-to-image generation. In International conference on machine learning . Pmlr, 8821–8831

    2021

  18. [26]

    Christos Sakaridis, Dengxin Dai, and Luc Van Gool. 2021. ACDC: The adverse conditions dataset with correspondences for semantic driving scene understand- ing. In Proceedings of the IEEE/CVF International Conference on Computer Vision . 10765–10775

    2021

  19. [27]

    Fengyi Shen, Li Zhou, Kagan Kucukaytekin, Ziyuan Liu, He Wang, and Alois Knoll. 2024. Controluda: Controllable diffusion-assisted unsupervised do- main adaptation for cross-weather semantic segmentation. arXiv preprint arXiv:2402.06446 (2024)

    2024 arXiv

  20. [28]

    Nataniel Ruiz, Yuanzhen Li, Varun Jampani, Yael Pritch, Michael Rubinstein, and Kfir Aberman. 2023. Dreambooth: Fine tuning text-to-image diffusion models for subject-driven generation. In Proceedings of the IEEE/CVF conference on computer vision and pattern recognition . 22500–22510

    2023

  21. [29]

    Jiapeng Su, Qi Fan, Wenjie Pei, Guangming Lu, and Fanglin Chen. 2024. Domain- rectifying adapter for cross-domain few-shot segmentation. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition . 24036–24045

    2024

  22. [30]

    Yuxi Wang, Jian Liang, Jun Xiao, Shuqi Mei, Yuran Yang, and Zhaoxiang Zhang

  23. [31]

    Inkyu Shin, Dong-Jin Kim, Jae Won Cho, Sanghyun Woo, Kwanyong Park, and In So Kweon. 2021. Labor: Labeling only if required for domain adaptive semantic segmentation. In Proceedings of the IEEE/CVF international conference on computer vision. 8588–8598

    2021

  24. [32]

    Pan Zhang, Bo Zhang, Ting Zhang, Dong Chen, Yong Wang, and Fang Wen. 2021. Prototypical pseudo label denoising and target structure learning for domain adaptive semantic segmentation. In Proceedings of the IEEE/CVF conference on computer vision and pattern recognition . 12414–12424

    2021

  25. [33]

    Kaiyang Zhou, Yongxin Yang, Yu Qiao, and Tao Xiang. 2021. Domain generaliza- tion with mixstyle. arXiv preprint arXiv:2104.02008 (2021)

    2021 arXiv

  26. [34]

    Jinjing Zhu, Haotian Bai, and Lin Wang. 2023. Patch-mix transformer for unsu- pervised domain adaptation: A game perspective. In Proceedings of the IEEE/CVF conference on computer vision and pattern recognition . 3561–3571

    2023

  27. [35]

    Senqiao Yang, Zhuotao Tian, Li Jiang, and Jiaya Jia. 2024. Unified Language- driven Zero-shot Domain Adaptation. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition . 23407–23415

    2024

  28. [39]

    Yang Zou, Zhiding Yu, BVK Kumar, and Jinsong Wang. 2018. Unsupervised domain adaptation for semantic segmentation via class-balanced self-training. In Proceedings of the European conference on computer vision (ECCV) . 289–305

    2018

  29. [2021]

    In International conference on machine learning

    Learning transferable visual models from natural language supervision. In International conference on machine learning . PMLR, 8748–8763

  30. [2023]

    In Proceedings of the IEEE/CVF International Conference on Computer Vision

    Informative data mining for one-shot cross-domain semantic segmentation. In Proceedings of the IEEE/CVF International Conference on Computer Vision. 1064– 1074

  31. [2025]

    In Proceedings of the Computer Vision and Pattern Recognition Conference

    VerbDiff: Text-Only Diffusion Models with Enhanced Interaction Aware- ness. In Proceedings of the Computer Vision and Pattern Recognition Conference . 8041–8050

Pith tools

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