Pith. sign in

REVIEW 4 major objections 6 minor 63 references

SPAST: Arbitrary Style Transfer with Style Priors via Pre-trained Large-scale Model

T0 review · 4 major / 6 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 Fast stylization reaches diffusion-level quality in 0.010s

desk verdict A plausible distillation framework with a believable speed story, but the quality claim rests on self-referential metrics and a missing preference score. read the letter →

arxiv 2505.08695 v1 pith:GOECS3JX submitted 2025-05-13 cs.CV

Zhanjie Zhang , Quanwei Zhang , Junsheng Luan , Mengyuan Yang , Yun Wang , Lei Zhao This is my paper · ORCID
classification cs.CV
keywords arbitrarystyletransferpriorlossLocal-globalWindowSizeStylizationModulelatentdiffusioncontentstructurepreservationreal-timeinferencedistillationWikiArt
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

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

The reading

This paper proposes SPAST, a style-transfer method that aims to combine the strengths of small-model and diffusion-based approaches: high-quality stylized output, faithful preservation of the content image's structure, and fast inference. The central claim, supported by experiments against several state-of-the-art methods, is that SPAST produces higher-quality stylized images than its competitors while running in about 0.010 seconds per image, a small fraction of the time needed by diffusion-based methods. The method works by distilling a style prior from a diffusion model fine-tuned on artistic images into a lightweight stylization network, so the expensive diffusion model is used only during training, not at inference. A sympathetic reader would care because this suggests the quality gains of large pre-trained models can be carried over to real-time style transfer.

What carries the argument

The central object is the Local-global Window Size Stylization Module (LGWSSM), a two-scale attention module that fuses style features into content features. The local window module blocks content and style features into $b \times b$ non-overlapping regions, matches each content region to its most similar style region, and applies attention-weighted mean and variance to transfer style statistics. The global window module then lets every content position attend to all style positions, covering cases where the closest local region fails to supply a matching style position. The second load-bearing piece is the style prior loss, which feeds the stylized image into a fixed Artistic Stable Diffusion and penalizes the difference between its predicted noise and the actual added noise, using a CLIP-based style embedding as the conditioning signal; this loss pulls the lightweight generator toward the diffusion model's notion of good stylization during training only.

What would settle it

Run SPAST on a held-out set of 50 styles far outside WikiArt (e.g., modern digital art, technical illustrations, and synthetic textures) and measure the style prior loss and human preference; if the loss stays high or outputs show artifacts or color drift on these styles, the generalized style-prior assumption fails.

Watch

Extended reading notes

Core claim

The paper's central discovery is that a small, fast arbitrary style transfer network can inherit the quality of a large pre-trained diffusion model through a style prior loss, without inheriting its inference cost or its tendency to distort content structure. Specifically, the authors fine-tune a latent diffusion model on WikiArt, condition it on a style image using a learned attention embedding from an image encoder, and use the denoising residual of this fixed Artistic Stable Diffusion as a training signal for the lightweight network. The lightweight network's own architecture combines a local window module, which matches content regions to the closest style regions, with a global window module, which lets each position attend to the full style image. In experiments, SPAST reports lower content loss, style loss, Gram loss, and LPIPS than several prior methods, and a user study gives SPAST a higher deception rate, while running at 0.010 seconds per 512×512 image.

Load-bearing premise

The quality improvement rests on the assumption that a diffusion model fine-tuned on WikiArt and conditioned on a style image via a learned attention embedding gives a reliable style prior for any new style image, including styles unlike anything in its training set.

Editorial extensions

If this is right

  • SPAST can stylize an arbitrary content-style pair in about 0.010 seconds, making high-quality style transfer practical for interactive or real-time applications.
  • The style prior loss can be attached to existing small-model style transfer methods to improve output quality without increasing inference time, as demonstrated by the paper on SaNet.
  • Content structure preservation is stronger than naive diffusion-based pipelines because the heavy diffusion network supervises only training, not inference.
  • The combination of local region matching and global attention in LGWSSM improves stylization quality over either module alone, according to the ablation study.
  • Training requires only unpaired content and style images from public datasets, so the method does not depend on paired stylized data.

Reading between the lines

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

  • The style prior loss is essentially a distillation recipe: any image-conditioned diffusion model could replace the fine-tuned network, so future improvements in style-conditioned diffusion priors would likely transfer directly to SPAST's outputs.
  • The same framework could be applied to other generative tasks where a small, fast generator is supervised by a large diffusion model, such as video style transfer or image-to-image translation, since the quality gain is mediated by the diffusion prior rather than by a larger feature extractor.
  • The ablation's sensitivity to time step $t=500$ suggests that a noise-level schedule or an adaptive time step could improve robustness on extreme styles, because the medium-noise stage is where both content and style patterns are balanced.
  • If the style prior generalizes only within WikiArt-like art, SPAST would likely need domain-specific fine-tuning for non-artistic styles; this boundary is not tested by the paper.
Share X Bluesky LinkedIn Reddit HN

Signed reviews

No signed human review yet.

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

4 major / 6 minor

Summary. The paper proposes SPAST, a two-stage arbitrary style transfer method. In stage one, a pretrained Stable Diffusion is fine-tuned on WikiArt to obtain an Artistic Stable Diffusion (ASD) that accepts a style image as conditioning via a CLIP image encoder and a learnable self-attention network. In stage two, a VGG-based encoder-decoder is trained with a novel Local-global Window Size Stylization Module (LGWSSM) and a style prior loss that distills the denoising knowledge of ASD into the small model. The paper claims that SPAST generates high-quality stylized images, preserves content structure, and runs at 0.010 s per image, and it reports comparisons with eight methods in terms of content/style loss, LPIPS, deception rate, preference score, and inference time.

Significance. The core idea of distilling style priors from a fine-tuned diffusion model into a fast feed-forward stylization network is promising and could be valuable beyond this specific architecture. If the claims hold, the 0.010 s inference time with quality comparable to diffusion-based methods would be a practical advance. The paper provides detailed loss formulations and implementation settings, and the qualitative examples are generally plausible. However, the quantitative evidence for quality superiority is weakened by the overlap between the reported metrics and SPAST's training losses, and the user-study reporting is incomplete. With a strengthened evaluation, the work could be a solid contribution to the style transfer literature.

major comments (4)
  1. [Section 3.2, Eq. (17)] The style prior loss L_sp is defined as a vector-valued expectation of the noise residual, not a scalar objective. A loss function must be a scalar to be minimized; the gradient in Eq. (18) is the score-distillation gradient of the squared norm E[ w(t) || eps_phi(z_t;...)-eps ||^2 ], but that squared norm is missing from Eq. (17). Please correct the definition or clarify explicitly that the vector form is an abuse of notation; otherwise the method is not reproducible as stated.
  2. [Section 4.3, Table 1] The quantitative evidence for the central quality claim is partly circular. The Content Loss and Style Loss reported in Table 1 are exactly the training objectives L_cont and L_sty defined in Eqs. (19) and (20), and the Gram Loss uses the same VGG feature statistics; the comparison methods were not trained on these objectives. The only independent evidence, the user study, is incomplete: the Preference Score for SPAST is missing in Table 1, and no confidence intervals or significance tests are reported. The paper should add an independent perceptual evaluation (e.g., a complete A/B study with statistics, or a style-specific metric not used in training) and report it fully.
  3. [Section 3.6, Fig. 6] The timestep t=500 in the style prior loss is selected by visually inspecting stylized outputs on the test set. This is a form of test-set tuning. Select t on a validation split and report test results with that fixed value, or explicitly disclose the selection procedure and how it affects the reported numbers.
  4. [Section 3.2] The style prior loss assumes that ASD, fine-tuned on WikiArt, can be conditioned on arbitrary style images through the CLIP image encoder and self-attention network. The paper provides no evidence that this conditioning generalizes to styles outside the WikiArt distribution; if conditioning fails on a novel style, the loss would give weak or misleading gradients. Please provide a qualitative or quantitative check of ASD's reconstruction or conditioning quality on held-out styles (e.g., style images not seen during fine-tuning).
minor comments (6)
  1. [Abstract] The phrase 'LGWSSM)tofuse' is missing a space; the text also contains several other typographical errors, such as inconsistent use of 'LMM' versus 'LLM' in the Related Work section and grammatical issues like 'high-quality stylized images'.
  2. [Section 3.3, Eq. (19)] The notation for the content loss sum is unclear: 'L∑ i=4' does not specify the layers. Please write the sum explicitly as a sum over ReLU4_1 and ReLU5_1.
  3. [Section 3.2, Eq. (15)] In Eq. (15), the noisy latent z_s_t is written as a function of the image I_s; it should be a function of the VAE-encoded latent z_s, as described in the text.
  4. [Section 4.3, Table 1] The Preference Score row has a '-' in the SPAST column, but the text reports that SPAST obtained 54.8% over DiffuseIT. Either fill in the value or explain the presentation.
  5. [Section 4.4, Fig. 7] The ablation table reports numerical values but no significance tests; it is unclear whether the differences between configurations are meaningful.
  6. [Section 2] The abbreviation 'LLM' for 'pre-trained large-scale model' is nonstandard and easily confused with 'large language model'; consider using 'LMM' consistently.

Circularity Check

1 steps flagged · score 5.0 of 10

Evaluation-on-training-loss overlap: Table 1's leading metrics are SPAST's own training objectives, and the Preference Score for SPAST is missing.

  1. fitted input called prediction [Section 3.3, Eqs. (19)-(20); Section 4.3, Table 1]
    "we use the content loss L_cont to constrain euclidean distance between content feature and stylized image feature ... (19) ... we use style loss L_s to constrain global style distribution between stylized images and style images ... (20) ... Content Loss ↓ 4.521 ... Style Loss ↓ 1.0928"

    The two headline metrics in Table 1 are the same functions minimized as training losses in Eq. (23): L = λ1 L_sty + λ2 L_cont + ... . SPAST is trained for 160,000 iterations to minimize exactly these objectives, so its low Content Loss (4.521) and Style Loss (1.0928) largely measure convergence to its own training objective, not independent stylization quality. The comparison is asymmetric because competing methods were not trained on SPAST's specific loss combination. Thus the main quantitative evidence for the 'high-quality stylized images' claim is, in part, the training loss reported as a prediction.

full rationale

The concrete circularity is the evaluation-on-training-objective overlap. Eq. (19) (content loss) and Eq. (20) (style loss) are explicitly included in the overall training objective Eq. (23), and Table 1 reports these same quantities as the primary quality metrics; low values for SPAST therefore partly restate that the optimizer minimized its own losses. This is a genuine partial circularity, though not a full one: LPIPS and Deception Rate are not training objectives, and the proposed LGWSSM module and style-prior loss are not defined in terms of the evaluation metrics. No load-bearing self-citation was found—the authors' prior works are cited as related context, not as the justification for the central derivation. A separate reporting defect, noted for completeness, is that the Preference Score row in Table 1 is '-' for SPAST while the text claims 54.8% preference over DiffuseIT; this removes the strongest independent human-preference evidence but is not itself a circular step. Overall, the central quality claim is only partly independent of the training procedure, so the circularity score is moderate.

Assumptions & free parameters 3 free parameters · 3 assumptions · 0 invented entities

The method rests on standard deep-learning assumptions (VGG feature semantics, diffusion noise stages) and on the paper-specific assumption that a fine-tuned Stable Diffusion can serve as a generalizable style prior. The main free parameters are loss weights and the timestep t=500, which is selected on test examples. No new physical or conceptual entities are introduced.

free parameters (3)
  • Timestep t for style prior loss = 500
    Selected in Sec. 3.6 by visually comparing t=100, 500, 1000 on test examples (Fig. 6). This hyperparameter controls the noise level and directly affects stylization quality; it is fitted to the examples shown, not set by a principled rule.
  • Loss weights λ1..λ5 = all 1
    Set to 1 in Sec. 4.1; weight sensitivity is analyzed in Fig. 8 (with λ4=5 and λ5=5 quality degrades), but the chosen values are not optimized on a held-out set.
  • Identity loss weights = λ_identity1=50, λ_identity2=1
    Adopted from SaNet [30] in Sec. 3.3, not tuned for SPAST.
assumptions (3)
  • domain assumption VGG feature layers encode content at deep levels (ReLU4_1, 5_1) and style at shallower levels (ReLU1_1 to 5_1)
    Used in Sec. 3.3 to define content and style losses; standard but an empirical assumption about feature semantics.
  • domain assumption Diffusion models have three noise stages: high noise affects structure, medium affects both, low affects style
    Invoked in Sec. 3.6 to justify setting t=500; cited from [48,1,56], but the mapping of mid-noise to style is heuristic.
  • ad hoc to paper The fine-tuned Artistic Stable Diffusion can be conditioned on a style image via a CLIP image encoder and learnable self-attention, and its denoising score provides a useful style prior for arbitrary content-style pairs
    Central to the style prior loss (Eq. 17); no direct evidence is given that the conditioning generalizes beyond WikiArt training styles.

how reviews work

0 comments
Cite this review

Pith. "Pith review of SPAST: Arbitrary Style Transfer with Style Priors via Pre-trained Large-scale Model." pith.science (2026). https://pith.science/paper/GOECS3JX

@misc{pith2026250508695,
  author       = {Pith},
  title        = {Pith review of: SPAST: Arbitrary Style Transfer with Style Priors via Pre-trained Large-scale Model},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/GOECS3JX}},
  note         = {Machine review of arXiv:2505.08695}
}
read the original abstract

Given an arbitrary content and style image, arbitrary style transfer aims to render a new stylized image which preserves the content image's structure and possesses the style image's style. Existing arbitrary style transfer methods are based on either small models or pre-trained large-scale models. The small model-based methods fail to generate high-quality stylized images, bringing artifacts and disharmonious patterns. The pre-trained large-scale model-based methods can generate high-quality stylized images but struggle to preserve the content structure and cost long inference time. To this end, we propose a new framework, called SPAST, to generate high-quality stylized images with less inference time. Specifically, we design a novel Local-global Window Size Stylization Module (LGWSSM)tofuse style features into content features. Besides, we introduce a novel style prior loss, which can dig out the style priors from a pre-trained large-scale model into the SPAST and motivate the SPAST to generate high-quality stylized images with short inference time.We conduct abundant experiments to verify that our proposed method can generate high-quality stylized images and less inference time compared with the SOTA arbitrary style transfer methods.

Figures

Figures reproduced from arXiv: 2505.08695 by the authors.

Figure 1
Figure 1. The first column shows the input images. The other three columns show the stylized images produced by SaNet [30], SaNet+SD [36], and SaNet+SP (Ours). 1. Introduction Arbitrary style transfer aims to render a stylized image which preserves the content image’s structure and possesses the style of the style image. Existing arbitrary style transfer methods can be generally divided into small model-based methods (SMM) an… view at source ↗
Figure 2
Figure 2. Overview of our SPAST, which consists of two stages. Stage one: finetuning a pre-trained Stable Diffusion, obtaining an Artistic Stable Diffusion. Stage two: Training an arbitrary style transfer model with style priors. decoder here follows the SaNet [30] and adopts the mirror structure of VGG. Argmax Reshuffle 3 2 4 1 1 3 2 4 Softmax Softmax Q K K Q Attention Score Attention Score [ × , × [PITH_FULL_IMAGE:figures/… view at source ↗
Figure 3
Figure 3. Illustration of Local-global Window Size Stylization Module (LGWSSM). 3.1. Local-global Window Size Stylization Module As shown in [PITH_FULL_IMAGE:figures/full_fig_p003_3.png] view at source ↗
Figures from the paper (5 more)
Figure 4
Figure 4. Figure 4: Qualitative comparisons. The 1 𝑠𝑡 col shows the input images. The other columns show the stylized images generated by other SOTA methods. Please zoom in for better observation. 𝑤(𝑡) is a scheduling coefficient, and we set it as 1 − ̄𝛼𝑡 . The lower sp means that the st…
Figure 5
Figure 5. Figure 5: Illustration about why the Local-global Window Size Stylization Module works. 3.6. The analysis of Timestep 𝑡 in 𝑠𝑝 Previous methods [48, 1, 56] have verified that diffusion models can be roughly divided into three stages based on noise levels: high, medium, and low. …
Figure 6
Figure 6. Figure 6: Stylized images by our proposed SPAST. It is worth noting that the timestep 𝑡 of 𝑠𝑝 is the key to affect the stylized image. We show stylized samples with different timestep 𝑡 (e.g., Timestep=100/500/1000). 4. Experiments 4.1. Implementation Details Stage One In this …
Figure 7
Figure 7. Figure 7: Ablation study. Content Style Ours More More More [PITH_FULL_IMAGE:figures/full_fig_p008_7.png]
Figure 8
Figure 8. Figure 8: An ablation study on how the choice of loss function affects the stylized images. Please zoom-in for better comparison. 57 hours). Similarly, SPAST is also trained for 160,000 iterations (costing 23 hours) in stage two. We also train other state-of-the-art (SOTA) metho…

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

63 extracted references · 48 canonical work pages

  1. [1]

    Anim- ageisworthmultiplewords:Multi-attributeinversionforconstrained text-to-image synthesis

    Agarwal,A.,Karanam,S.,Shukla,T.,Srinivasan,B.V.,2023. Anim- ageisworthmultiplewords:Multi-attributeinversionforconstrained text-to-image synthesis. arXiv preprint arXiv:2311.11919

    arXiv 2023

  2. [2]

    Artflow: Unbiasedimagestyletransferviareversibleneuralflows,in:Proceed- ings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pp

    An, J., Huang, S., Song, Y., Dou, D., Liu, W., Luo, J., 2021. Artflow: Unbiasedimagestyletransferviareversibleneuralflows,in:Proceed- ings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pp. 862–871

    work page 2021

  3. [3]

    Relactrl:Relevance-guidedefficient controlfordiffusiontransformers

    Cao,K.,Wang,J.,Ma,A.,Feng,J.,Zhang,Z.,He,X.,Liu,S.,Cheng, B.,Leng,D.,Yin,Y.,etal.,2025. Relactrl:Relevance-guidedefficient controlfordiffusiontransformers. arXivpreprintarXiv:2502.14377

    arXiv 2025

  4. [4]

    Artistic style transfer with internal-external learning andcontrastivelearning

    Chen, H., Wang, Z., Zhang, H., Zuo, Z., Li, A., Xing, W., Lu, D., et al., 2021a. Artistic style transfer with internal-external learning andcontrastivelearning. AdvancesinNeuralInformationProcessing Systems 34, 26561–26573

    work page

  5. [5]

    Dualast:Dualstyle-learningnetworksforartisticstyle transfer, in: Proceedings of the IEEE/CVF conference on computer vision and pattern recognition, pp

    Chen, H., Zhao, L., Wang, Z., Zhang, H., Zuo, Z., Li, A., Xing, W., Lu,D.,2021b. Dualast:Dualstyle-learningnetworksforartisticstyle transfer, in: Proceedings of the IEEE/CVF conference on computer vision and pattern recognition, pp. 872–881

    work page

  6. [6]

    Fast patch-based style transfer of arbitrary style

    Chen, T.Q., Schmidt, M., 2016. Fast patch-based style transfer of arbitrary style. arXiv preprint arXiv:1612.04337

    arXiv 2016

  7. [7]

    General image-to-image translation with one-shot image guidance, in: Proceedings of the IEEE/CVFInternationalConferenceonComputerVision,pp.22736– 22746

    Cheng, B., Liu, Z., Peng, Y., Lin, Y., 2023. General image-to-image translation with one-shot image guidance, in: Proceedings of the IEEE/CVFInternationalConferenceonComputerVision,pp.22736– 22746

    work page 2023

  8. [8]

    Z*: Zero-shot style transfer via attention reweighting, in: Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pp

    Deng, Y., He, X., Tang, F., Dong, W., 2024. Z*: Zero-shot style transfer via attention reweighting, in: Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pp. 6934– 6944

    work page 2024

Show all 63 references
  1. [9]

    Stytr2: Image style transfer with transformers, in: Proceedings of the IEEE/CVFConferenceonComputerVisionandPatternRecognition, pp

    Deng,Y.,Tang,F.,Dong,W.,Ma,C.,Pan,X.,Wang,L.,Xu,C.,2022. Stytr2: Image style transfer with transformers, in: Proceedings of the IEEE/CVFConferenceonComputerVisionandPatternRecognition, pp. 11326–11336

    2022

  2. [10]

    Arbitrarystyletransferviamulti-adaptationnetwork,in:Proceedings of the 28th ACM international conference on multimedia, pp

    Deng, Y., Tang, F., Dong, W., Sun, W., Huang, F., Xu, C., 2020. Arbitrarystyletransferviamulti-adaptationnetwork,in:Proceedings of the 28th ACM international conference on multimedia, pp. 2719– 2727

    2020

  3. [11]

    Diffusion models beat gans on image synthesis

    Dhariwal, P., Nichol, A., 2021. Diffusion models beat gans on image synthesis. Advances in Neural Information Processing Systems 34, 8780–8794

    2021

  4. [12]

    Styleshot: A snapshot on any style

    Gao,J.,Liu,Y.,Sun,Y.,Tang,Y.,Zeng,Y.,Chen,K.,Zhao,C.,2024. Styleshot: A snapshot on any style. arXiv preprint arXiv:2407.01414

    2024 arXiv

  5. [13]

    A neural algorithm of artistic style

    Gatys, L.A., Ecker, A.S., Bethge, M., 2015. A neural algorithm of artistic style. arXiv preprint arXiv:1508.06576

    2015 arXiv

  6. [14]

    Gulrajani,I.,Ahmed,F.,Arjovsky,M.,Dumoulin,V.,Courville,A.C.,

  7. [15]

    Cartoondiff: Training-free cartoon image generation with diffusion transformer models

    He, F., Li, G., Si, L., Yan, L., Hou, S., Dong, H., Li, F., 2023. Cartoondiff: Training-free cartoon image generation with diffusion transformer models. arXiv preprint arXiv:2309.08251

    2023 arXiv

  8. [16]

    Denoising diffusion probabilistic models

    Ho, J., Jain, A., Abbeel, P., 2020. Denoising diffusion probabilistic models. Advances in neural information processing systems 33, 6840–6851

    2020

  9. [17]

    Aespa-net: Aesthetic pattern-aware style transfer networks,in:ProceedingsoftheIEEE/CVFInternationalConference on Computer Vision, pp

    Hong, K., Jeon, S., Lee, J., Ahn, N., Kim, K., Lee, P., Kim, D., Uh, Y., Byun, H., 2023. Aespa-net: Aesthetic pattern-aware style transfer networks,in:ProceedingsoftheIEEE/CVFInternationalConference on Computer Vision, pp. 22758–22767

    2023

  10. [18]

    Arbitrary style transfer in real-time with adaptive instance normalization, in: 2017 IEEE International Conference on Computer Vision (ICCV), pp

    Huang, X., Belongie, S., 2017. Arbitrary style transfer in real-time with adaptive instance normalization, in: 2017 IEEE International Conference on Computer Vision (ICCV), pp. 1510–1519

    2017

  11. [19]

    Uncorrelated feature encoding for faster image style transfer

    Kim, M., Choi, H.C., 2021. Uncorrelated feature encoding for faster image style transfer. Neural Networks 140, 148–157

    2021

  12. [20]

    Adam:Amethodforstochasticoptimiza- tion

    Kingma,D.P.,Ba,J.,2014. Adam:Amethodforstochasticoptimiza- tion. arXiv preprint arXiv:1412.6980

    2014 arXiv

  13. [21]

    Diffusion-based image translation using disentangled style and content representation, in: The Eleventh Inter- national Conference on Learning Representations

    Kwon, G., Ye, J.C., 2023. Diffusion-based image translation using disentangled style and content representation, in: The Eleventh Inter- national Conference on Learning Representations

    2023

  14. [22]

    Li, G., Zhao, L., Sun, J., Lan, Z., Zhang, Z., Chen, J., Lin, Z., Lin, H., Xing, W., 2023. Rethinking multi-contrast mri super-resolution: Rectangle-windowcross-attentiontransformerandarbitrary-scaleup- sampling,in:ProceedingsoftheIEEE/CVFInternationalConference on Computer Vi...

    2023

  15. [23]

    Tuigan: Learning versatile image-to-image translation with two unpaired images, in: European Conference on Computer Vision, Springer

    Lin, J., Pang, Y., Xia, Y., Chen, Z., Luo, J., 2020. Tuigan: Learning versatile image-to-image translation with two unpaired images, in: European Conference on Computer Vision, Springer. pp. 18–35

    2020

  16. [24]

    Microsoft coco: Common objects in context, in: European conference on computer vision, Springer

    Lin,T.Y.,Maire,M.,Belongie,S.,Hays,J.,Perona,P.,Ramanan,D., Dollár, P., Zitnick, C.L., 2014. Microsoft coco: Common objects in context, in: European conference on computer vision, Springer. pp. 740–755

    2014

  17. [25]

    Intrinsic-style distribution matching for arbitrary style transfer

    Liu, M., Lin, S., Zhang, H., Zha, Z., Wen, B., 2024. Intrinsic-style distribution matching for arbitrary style transfer. Knowledge-Based Systems 296, 111898

    2024

  18. [26]

    Adaattn: Revisit attention mechanism in arbitrary neural style transfer, in: Proceedings of the IEEE/CVF international conference on computer vision, pp

    Liu, S., Lin, T., He, D., Li, F., Wang, M., Li, X., Sun, Z., Li, Q., Ding, E., 2021. Adaattn: Revisit attention mechanism in arbitrary neural style transfer, in: Proceedings of the IEEE/CVF international conference on computer vision, pp. 6649–6658

    2021

  19. [27]

    Image style transfer with collection representation space and semantic-guided reconstruction

    Ma, Z., Li, J., Wang, N., Gao, X., 2020. Image style transfer with collection representation space and semantic-guided reconstruction. Neural Networks 129, 123–137

    2020

  20. [28]

    Improveddenoisingdiffusionprob- abilistic models, in: International Conference on Machine Learning, PMLR

    Nichol,A.Q.,Dhariwal,P.,2021. Improveddenoisingdiffusionprob- abilistic models, in: International Conference on Machine Learning, PMLR. pp. 8162–8171

    2021

  21. [29]

    Painter by numbers, wikiart

    Nichol, K., 2016. Painter by numbers, wikiart

    2016

  22. [30]

    Arbitrary style transfer with style- attentional networks, in: Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition, pp

    Park, D.Y., Lee, K.H., 2019. Arbitrary style transfer with style- attentional networks, in: Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition, pp. 5880–5888

    2019

  23. [31]

    Parmar, G., Singh, K.K., Zhang, R., Li, Y., Lu, J., Zhu, J.Y.,

  24. [32]

    Pytorch:Animperativestyle,high-performancedeeplearninglibrary

    Paszke, A., Gross, S., Massa, F., Lerer, A., Bradbury, J., Chanan, G., Killeen, T., Lin, Z., Gimelshein, N., Antiga, L., et al., 2019. Pytorch:Animperativestyle,high-performancedeeplearninglibrary. Advances in neural information processing systems 32

    2019

  25. [33]

    Dreamfusion: Text-to-3d using 2d diffusion, in: The Eleventh International Confer- ence on Learning Representations

    Poole, B., Jain, A., Barron, J.T., Mildenhall, B., 2022. Dreamfusion: Text-to-3d using 2d diffusion, in: The Eleventh International Confer- ence on Learning Representations. Z. Zhang et al.:Preprint submitted to Elsevier Page 9 of 10 SPAST

    2022

  26. [34]

    Learning transferable visual models from natural language supervision, in: International Conference on Machine Learning, PMLR

    Radford, A., Kim, J.W., Hallacy, C., Ramesh, A., Goh, G., Agarwal, S.,Sastry,G.,Askell,A.,Mishkin,P.,Clark,J.,etal.,2021. Learning transferable visual models from natural language supervision, in: International Conference on Machine Learning, PMLR. pp. 8748– 8763

    2021

  27. [35]

    Unsupervised repre- sentation learning with deep convolutional generative adversarial networks

    Radford, A., Metz, L., Chintala, S., 2015. Unsupervised repre- sentation learning with deep convolutional generative adversarial networks. arXiv preprint arXiv:1511.06434

    2015 arXiv

  28. [36]

    High-resolution image synthesis with latent diffusion models, in: Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pp

    Rombach,R.,Blattmann,A.,Lorenz,D.,Esser,P.,Ommer,B.,2022. High-resolution image synthesis with latent diffusion models, in: Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pp. 10684–10695

    2022

  29. [37]

    Very deep convolutional networks for large-scale image recognition

    Simonyan, K., Zisserman, A., 2014. Very deep convolutional networks for large-scale image recognition. arXiv preprint arXiv:1409.1556

    2014 arXiv

  30. [38]

    Attention is all you need

    Vaswani,A.,Shazeer,N.,Parmar,N.,Uszkoreit,J.,Jones,L.,Gomez, A.N., Kaiser, Ł., Polosukhin, I., 2017. Attention is all you need. Advances in neural information processing systems 30

    2017

  31. [39]

    Fine- grained image style transfer with visual transformers

    Wang, J., Yang, H., Fu, J., Yamasaki, T., Guo, B., 2022a. Fine- grained image style transfer with visual transformers. arXiv preprint arXiv:2210.05176

    arXiv

  32. [40]

    Aesust: Towards aesthetic-enhanced universal style transfer, in: Proceedings of the 30th ACM International Conference on Multi- media, pp

    Wang, Z., Zhang, Z., Zhao, L., Zuo, Z., Li, A., Xing, W., Lu, D., 2022b. Aesust: Towards aesthetic-enhanced universal style transfer, in: Proceedings of the 30th ACM International Conference on Multi- media, pp. 1095–1106

  33. [41]

    Stylediffusion: Controllable disentangledstyletransferviadiffusionmodels,in:Proceedingsofthe IEEE/CVF International Conference on Computer Vision, pp

    Wang, Z., Zhao, L., Xing, W., 2023. Stylediffusion: Controllable disentangledstyletransferviadiffusionmodels,in:Proceedingsofthe IEEE/CVF International Conference on Computer Vision, pp. 7677– 7689

    2023

  34. [42]

    Cycle consistent network for end-to-end style transfer tts training

    Xue, L., Pan, S., He, L., Xie, L., Soong, F.K., 2021. Cycle consistent network for end-to-end style transfer tts training. Neural Networks 140, 223–236

    2021

  35. [43]

    Gating pat- ternpyramidfordiversifiedimagestyletransfer

    Yang, F., Chen, H., Zhang, Z., Zhao, L., Lin, H., 2022. Gating pat- ternpyramidfordiversifiedimagestyletransfer. JournalofElectronic Imaging 31, 063007

    2022

  36. [44]

    Zero-shot contrastive loss for text-guided diffusion image style transfer

    Yang, S., Hwang, H., Ye, J.C., 2023. Zero-shot contrastive loss for text-guided diffusion image style transfer

    2023

  37. [45]

    Aesstyler: Aesthetic guided universal style transfer, in: Proceedings of the 32nd ACM International Conference on Multimedia, pp

    Yi, R., Zhu, H., Hu, T., Lai, Y.K., Rosin, P.L., 2024. Aesstyler: Aesthetic guided universal style transfer, in: Proceedings of the 32nd ACM International Conference on Multimedia, pp. 9789–9798

    2024

  38. [46]

    Dif-fusion:Towards high color fidelity in infrared and visible image fusion with diffusion models

    Yue,J.,Fang,L.,Xia,S.,Deng,Y.,Ma,J.,2023. Dif-fusion:Towards high color fidelity in infrared and visible image fusion with diffusion models. arXiv preprint arXiv:2301.08072

    2023 arXiv

  39. [47]

    The unreasonableeffectivenessofdeepfeaturesasaperceptualmetric,in: Proceedings of the IEEE conference on computer vision and pattern recognition, pp

    Zhang,R.,Isola,P.,Efros,A.A.,Shechtman,E.,Wang,O.,2018. The unreasonableeffectivenessofdeepfeaturesasaperceptualmetric,in: Proceedings of the IEEE conference on computer vision and pattern recognition, pp. 586–595

    2018

  40. [48]

    Prospect: Prompt spectrum for attribute-aware personalization of diffusion models

    Zhang, Y., Dong, W., Tang, F., Huang, N., Huang, H., Ma, C., Lee, T.Y., Deussen, O., Xu, C., 2023a. Prospect: Prompt spectrum for attribute-aware personalization of diffusion models. ACM Transac- tions on Graphics (TOG) 42, 1–14

  41. [49]

    Inversion-based creativity transfer with diffusion models

    Zhang, Y., Huang, N., Tang, F., Huang, H., Ma, C., Dong, W., Xu, C., 2023b. Inversion-based creativity transfer with diffusion models. Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition

  42. [50]

    Zhang, Y., Tang, F., Dong, W., Huang, H., Ma, C., Lee, T.Y., Xu, C.,

  43. [51]

    Csast: Content self-supervised and style contrastive learning for arbitrary style transfer

    Zhang, Y., Tian, Y., Hou, J., 2023c. Csast: Content self-supervised and style contrastive learning for arbitrary style transfer. Neural Networks 164, 146–155

  44. [52]

    Neurocomputing 623, 129434

    Zhang, Z., Li, Y., Xia, R., Yang, M., Wang, Y., Zhao, L., Xing, W., 2025a.Lgast:Towardshigh-qualityarbitrarystyletransferwithlocal– global style learning. Neurocomputing 623, 129434

  45. [53]

    Caster: Cartoon style transfer via dynamic cartoon style casting

    Zhang,Z.,Sun,J.,Chen,J.,Zhao,L.,Ji,B.,Lan,Z.,Li,G.,Xing,W., Xu, D., 2023d. Caster: Cartoon style transfer via dynamic cartoon style casting. Neurocomputing 556, 126654

  46. [54]

    Rethink arbitrary style transfer with transformer and contrastive learning

    Zhang,Z.,Sun,J.,Li,G.,Zhao,L.,Zhang,Q.,Lan,Z.,Yin,H.,Xing, W., Lin, H., Zuo, Z., 2024a. Rethink arbitrary style transfer with transformer and contrastive learning. Computer Vision and Image Understanding 241, 103951

  47. [55]

    Dyartbank: Diverse artistic style transfer via pre-trained stablediffusionanddynamicstylepromptartbank

    Zhang, Z., Zhang, Q., Li, G., Luan, J., Yang, M., Wang, Y., Zhao, L., 2025b. Dyartbank: Diverse artistic style transfer via pre-trained stablediffusionanddynamicstylepromptartbank. Knowledge-Based Systems 310, 112959

  48. [56]

    Zhang, Z., Zhang, Q., Lin, H., Xing, W., Mo, J., Huang, S., Xie, J., Li, G., Luan, J., Zhao, L., et al., 2024b. Towards highly realistic artistic style transfer via stable diffusion with step-aware and layer- aware prompt, in: Proceedings of the Thirty-Third International Join...

  49. [57]

    Artbank: Artistic style transfer with pre-trained diffusion model and implicit style prompt bank, in: Proceedings of the AAAI Conference on Artificial Intelligence, pp

    Zhang, Z., Zhang, Q., Xing, W., Li, G., Zhao, L., Sun, J., Lan, Z., Luan, J., Huang, Y., Lin, H., 2024c. Artbank: Artistic style transfer with pre-trained diffusion model and implicit style prompt bank, in: Proceedings of the AAAI Conference on Artificial Intelligence, pp. 7396–7404

  50. [58]

    Unpaired image-to-image transla- tion using adversarial consistency loss, in: European Conference on Computer Vision, Springer

    Zhao, Y., Wu, R., Dong, H., 2020. Unpaired image-to-image transla- tion using adversarial consistency loss, in: European Conference on Computer Vision, Springer. pp. 800–815

    2020

  51. [59]

    Unpaired image- to-image translation using cycle-consistent adversarial networks, in: ProceedingsoftheIEEEinternationalconferenceoncomputervision, pp

    Zhu, J.Y., Park, T., Isola, P., Efros, A.A., 2017. Unpaired image- to-image translation using cycle-consistent adversarial networks, in: ProceedingsoftheIEEEinternationalconferenceoncomputervision, pp. 2223–2232

    2017

  52. [60]

    All-to-key attentionforarbitrarystyletransfer,in:ProceedingsoftheIEEE/CVF International Conference on Computer Vision, pp

    Zhu, M., He, X., Wang, N., Wang, X., Gao, X., 2023. All-to-key attentionforarbitrarystyletransfer,in:ProceedingsoftheIEEE/CVF International Conference on Computer Vision, pp. 23109–23119. Z. Zhang et al.:Preprint submitted to Elsevier Page 10 of 10

    2023

  53. [2017]

    5767–5777

    Improved training of wasserstein gans, in: Advances in neural information processing systems, pp. 5767–5777

  54. [2022]

    arXiv preprint arXiv:2205.09542

    Domainenhancedarbitraryimagestyletransferviacontrastive learning. arXiv preprint arXiv:2205.09542

    arXiv

  55. [2023]

    arXiv preprint arXiv:2302.03027

    Zero-shot image-to-image translation. arXiv preprint arXiv:2302.03027

    arXiv

Pith tools

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