Pith. sign in

REVIEW 5 major objections 3 minor 63 references

Undress to Redress: A Training-Free Framework for Virtual Try-On

T0 review · 5 major / 3 minor · reviewed 2026-08-05 · deepseek-v4-flash

Pith's one-line read Undress-to-redress: a training-free framework that improves long-sleeve-to-short-sleeve virtual try-on by first reconstructing a bare torso, then applying the target garment.

desk verdict Plausible two-stage idea for long-to-short-sleeve try-on, but with the full text unreadable, all load-bearing claims rest on an abstract that gives no evidence for its core assumption. read the letter →

arxiv 2508.07680 v1 pith:AWJM6TRJ submitted 2025-08-11 cs.CV

classification cs.CV
keywords virtualtry-onlong-sleeve-to-short-sleeveundress-to-redresstraining-freeframeworkdiffusionmodelsclassifier-freeguidancebenchmarkimagegeneration
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

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

The reading

This paper argues that virtual try-on models fail at long-sleeve-to-short-sleeve conversions because they follow a 'majority' completion rule: when part of the arm is hidden by the original sleeve, the model fills the exposed skin by copying the dominant clothing texture instead of reconstructing skin. The authors propose UR-VTON, a training-free wrapper that first 'undresses' the person to produce a bare-torso image and then applies the target short-sleeve garment, decomposing the one-step conversion into two simpler steps. They also add Dynamic Classifier-Free Guidance scheduling during diffusion sampling and a Structural Refiner that uses high-frequency image cues to preserve detail. The paper introduces a new benchmark, LS-TON, for this scenario and reports that UR-VTON beats state-of-the-art methods on detail preservation and image quality. If correct, any existing VTON method could be upgraded for short-sleeve try-on without retraining.

What carries the argument

The load-bearing mechanism is the undress-to-redress decomposition: first a virtual undressing pass reconstructs a bare-torso image of the user, then the target short-sleeve garment is applied to that torso. Two auxiliaries carry the quality claim: Dynamic Classifier-Free Guidance scheduling (varying the guidance scale during the denoising steps of a diffusion model to balance diversity and fidelity) and a Structural Refiner (a detail-enhancement pass driven by high-frequency image cues).

What would settle it

Run UR-VTON on LS-TON with the undress stage replaced by a straight copy of the input (or by a trivial identity), holding the CFG schedule and Structural Refiner fixed. If the final short-sleeve images are not measurably worse in detail preservation and perceptual quality than the full pipeline, the undress-to-redress mechanism is not what carries the improvement. Alternatively, if a controlled user study on LS-TON finds no preference for UR-VTON over direct one-stage methods, the central claim fails.

Watch

Extended reading notes

Core claim

The central claim is that the reason long-sleeve-to-short-sleeve try-on images look wrong is not a missing training signal but a completion bias: diffusion-based VTON models, when asked to expose arm skin that appears nowhere in the input, default to the 'majority' appearance of the neighboring pixels—the sleeve—producing smeared or fabric-textured skin. UR-VTON counteracts this by introducing an intermediate 'undress' step that reconstructs a plausible bare-torso image from the clothed photo, so that the subsequent 'redress' step only needs to place a short-sleeve garment onto an already plausible body. Because the framework is training-free, it can wrap any existing VTON pipeline. On top o

Load-bearing premise

The framework's central bet is that a model can first produce a believable bare-torso image of the person (correct skin, identity, and geometry) and that this intermediate makes the final short-sleeve image easier to generate than going directly from long sleeves.

Editorial extensions

If this is right

  • Any existing VTON method can adopt UR-VTON without retraining, so long-sleeve-to-short-sleeve capability becomes a plug-in upgrade rather than a new model.
  • Long-to-short-sleeve try-on outputs should show sharper, more skin-realistic arms than direct one-stage conversion, which the paper presents as evidence for the 'majority completion rule' diagnosis.
  • The LS-TON benchmark gives the community a standard set of long-to-short-sleeve cases to measure detail preservation and image quality.
  • The Dynamic CFG schedule and Structural Refiner are claimed to generalize beyond the undress step, improving sampling balance and detail in any diffusion-based try-on pipeline.

Reading between the lines

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

  • If undress-to-redress works because it resets the 'majority' bias, the same trick should transfer to other skin-exposing edits—short-sleeve to sleeveless, long pants to shorts—where exposed skin is absent from the input; the paper does not test these.
  • The framework's success hinges on the undress stage being accurate; a natural stress test is to swap in different undress models and see whether final quality tracks undress fidelity, which the paper does not isolate.
  • One could also invert the pipeline to redress-to-undress for garment-removal applications; that direction is outside the paper's stated scope.
Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

5 major / 3 minor

Summary. The paper proposes UR-VTON, a training-free framework for long-sleeve-to-short-sleeve virtual try-on. The claimed mechanism is a two-stage decomposition: first 'undress' the user by generating a bare-torso image, then 'redress' with the target short-sleeve garment. Two additional components are introduced: Dynamic Classifier-Free Guidance scheduling and a Structural Refiner for detail fidelity. The paper also introduces a new benchmark, LS-TON, for this task. The abstract states that extensive experiments demonstrate superiority over state-of-the-art methods in detail preservation and image quality. The supplied full text, however, is severely corrupted: equations, tables, and most prose are unreadable glyphs, so none of the technical details, experimental protocols, or quantitative results can be verified from the manuscript as submitted.

Significance. If the claims hold, the undress-to-redress decomposition is a conceptually interesting and practically useful alternative to direct one-stage conversion, and the training-free integration with existing VTON methods would be a valuable contribution. A dedicated benchmark for long-sleeve-to-short-sleeve try-on could also fill a real gap. However, because the manuscript text is largely unreadable and the abstract provides only unsupported high-level claims, the significance cannot currently be grounded in verifiable evidence. The paper would need a readable, complete version with rigorous evaluation before its contribution can be assessed.

major comments (5)
  1. [Full text (general)] The submitted manuscript is almost entirely unreadable due to encoding corruption. Equations, tables, figure captions, and method descriptions are garbled, so no implementation detail, ablation, or numerical result can be checked. This alone blocks acceptance: a referee cannot verify the central claim. The authors must resubmit a properly encoded PDF or source. I cannot recommend acceptance or even a focused revision without being able to read the technical content.
  2. [Undress stage (method)] The framework's load-bearing assumption is that virtually 'undressing' the user produces an accurate bare-torso intermediate, and that this intermediate is easier to dress than to convert directly. The abstract provides no evaluation of the intermediate undress outputs—no skin fidelity, identity preservation, or body-shape consistency metrics. If the undress stage hallucinates or distorts skin, those errors propagate to the final short-sleeve result. The authors should report quantitative and qualitative evaluation of the undress stage itself, and show that final output quality is not systematically worse when the undress stage is imperfect.
  3. [Abstract / motivation] The 'majority completion rule' is stated as the cause of long-sleeve-to-short-sleeve failures, but no evidence or formal definition is provided in the readable portions of the paper. This is an ad-hoc diagnostic hypothesis, not a demonstrated mechanism. The authors should define the rule operationally, validate it against existing VTON models, and show that the undress-to-redress decomposition specifically addresses it rather than merely circumventing it.
  4. [Experiments (abstract)] The abstract claims 'extensive experiments' and superiority over state-of-the-art methods without naming baselines, datasets, or metrics, and without error bars or statistical tests. Even if the full text were readable, the abstract's level of specificity is insufficient for a central empirical claim. The authors should specify the compared methods, the LS-TON construction and size, the evaluation metrics, and report variance or significance tests across runs or subjects.
  5. [Dynamic CFG scheduling / Structural Refiner] The two additional components, Dynamic Classifier-Free Guidance scheduling and Structural Refiner, are introduced as key parts of the framework, but the readable text provides no equations, algorithm descriptions, or ablation results for either. The authors should provide precise definitions and controlled ablations showing the contribution of each component to the final result, including the schedule's free parameters and sensitivity.
minor comments (3)
  1. [Abstract] The phrase 'majority completion rule' appears in quotes but is not defined in the abstract; consider a brief informal explanation or a reference to the section where it is formalized.
  2. [General] The paper promises code release upon acceptance; this is welcome, but the current manuscript contains no reproducibility artifact or pseudocode that could be checked even if the text were readable.
  3. [General] If the encoding issue is an artifact of the arXiv source, the authors should ensure the PDF and source are both cleanly encoded; the current version appears as a sequence of replacement characters, making even section headings ambiguous.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: the proposed UR-VTON pipeline is an inference-time composition around a pretrained VTON model, with no fitted parameter renamed as a prediction and no load-bearing self-citation chain.

full rationale

I examined the claimed derivation chain: (1) the 'majority completion rule' diagnosis, (2) the undress-to-redress decomposition, (3) Dynamic CFG scheduling, (4) Structural Refiner, and (5) the LS-TON benchmark. None of these are shown to be equivalent to their inputs by construction. The method is training-free and is presented as a plug-in wrapper around existing VTON methods, so there is no fitted parameter that is later relabeled as a prediction. No equation-level self-definitional step can be identified from the available text; the equations present in the manuscript are too garbled to support any claim such as 'Eq. X = Eq. Y by construction.' The paper does introduce its own benchmark LS-TON, but introducing and evaluating on a new benchmark is not circular by itself: no evidence in the text shows that the benchmark was constructed so that UR-VTON's output is forced by the benchmark definition. The main substantive weakness is that the undress stage's reliability is assumed rather than demonstrated, and intermediate bare-torso outputs are not evaluated; however, that is a correctness/robustness concern, not a circularity concern. There is also no visible load-bearing self-citation: the manuscript's core mechanism is described as a new inference-time framework rather than as a consequence of prior work by the same authors. Therefore the appropriate finding is no significant circularity, score 0.

Assumptions & free parameters 1 free parameters · 3 assumptions · 0 invented entities

The central claims rest on the existence of a competent pretrained VTON model (training-free wrapper), on the unproven 'majority completion rule' diagnosis, and on the assumption that a virtual undressing intermediate is accurate. Because the full text is corrupted, additional implicit assumptions in the diffusion sampling and structural refinement modules could not be audited.

free parameters (1)
  • Dynamic CFG schedule shape = not specified in abstract
    Dynamic Classifier-Free Guidance scheduling is a stated component; its schedule (guidance weight as a function of sampling timestep) is a design choice not described in the abstract and not checkable in the corrupted full text.
assumptions (3)
  • domain assumption A pretrained VTON model exists and is fixed; UR-VTON needs no training
    The entire training-free claim depends on an underlying VTON method that already performs reasonable garment transfer, so that a two-stage wrapper improves it.
  • ad hoc to paper The 'majority completion rule' explains long-to-short-sleeve failures
    The abstract asserts that current VTON models follow a majority completion rule leading to inaccurate skin restoration; this is a diagnosis offered without proof or ablation in the readable text.
  • domain assumption Virtual undressing is a reliable first step that preserves identity and body geometry
    The undress-to-redress mechanism assumes the undressed intermediate image is accurate enough that the subsequent dressing step is easier than direct conversion; if undressing introduces artifacts, the pipeline inherits them.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Undress to Redress: A Training-Free Framework for Virtual Try-On." pith.science (2026). https://pith.science/paper/AWJM6TRJ

@misc{pith2026250807680,
  author       = {Pith},
  title        = {Pith review of: Undress to Redress: A Training-Free Framework for Virtual Try-On},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/AWJM6TRJ}},
  note         = {Machine review of arXiv:2508.07680}
}
read the original abstract

Virtual try-on (VTON) is a crucial task for enhancing user experience in online shopping by generating realistic garment previews on personal photos. Although existing methods have achieved impressive results, they struggle with long-sleeve-to-short-sleeve conversions-a common and practical scenario-often producing unrealistic outputs when exposed skin is underrepresented in the original image. We argue that this challenge arises from the ''majority'' completion rule in current VTON models, which leads to inaccurate skin restoration in such cases. To address this, we propose UR-VTON (Undress-Redress Virtual Try-ON), a novel, training-free framework that can be seamlessly integrated with any existing VTON method. UR-VTON introduces an ''undress-to-redress'' mechanism: it first reveals the user's torso by virtually ''undressing,'' then applies the target short-sleeve garment, effectively decomposing the conversion into two more manageable steps. Additionally, we incorporate Dynamic Classifier-Free Guidance scheduling to balance diversity and image quality during DDPM sampling, and employ Structural Refiner to enhance detail fidelity using high-frequency cues. Finally, we present LS-TON, a new benchmark for long-sleeve-to-short-sleeve try-on. Extensive experiments demonstrate that UR-VTON outperforms state-of-the-art methods in both detail preservation and image quality. Code will be released upon acceptance.

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

63 extracted references · 52 canonical work pages

  1. [1]

    , " * write output.state after.block = add.period write newline

    ENTRY address archivePrefix author booktitle chapter edition editor eid eprint howpublished institution isbn journal key month note number organization pages publisher school series title type volume year label extra.label sort.label short.list INTEGERS output.state before.all mid.sentence after.sentence after.block FUNCTION init.state.consts #0 'before.a...

  2. [2]

    write newline

    " write newline "" before.all 'output.state := FUNCTION n.dashify 't := "" t empty not t #1 #1 substring "-" = t #1 #2 substring "--" = not "--" * t #2 global.max substring 't := t #1 #1 substring "-" = "-" * t #2 global.max substring 't := while if t #1 #1 substring * t #2 global.max substring 't := if while FUNCTION word.in bbl.in capitalize " " * FUNCT...

  3. [3]

    Bai, S.; Zhou, H.; Li, Z.; Zhou, C.; and Yang, H. 2022. Single stage virtual try-on via deformable attention flows. In European Conference on Computer Vision, 409--425

  4. [4]

    Baldrati, A.; Morelli, D.; Cartella, G.; Cornia, M.; Bertini, M.; and Cucchiara, R. 2023. Multimodal garment designer: Human-centric latent diffusion models for fashion image editing. In International Conference on Computer Vision, 23393--23402

  5. [5]

    J.; Arbel, M.; and Gretton, A

    Bi \'n kowski, M.; Sutherland, D. J.; Arbel, M.; and Gretton, A. 2018. Demystifying mmd gans. arXiv preprint arXiv:1801.01401

  6. [6]

    Bookstein, F. L. 2002. Principal warps: Thin-plate splines and the decomposition of deformations. IEEE Transactions on Pattern Analysis and Machine Intelligence, 11(6): 567--585

  7. [7]

    Chen, J.; Duanmu, C.; and Long, H. 2024. Large kernel frequency-enhanced network for efficient single image super-resolution. In Computer Vision and Pattern Recognition, 6317--6326

  8. [8]

    Choi, S.; Park, S.; Lee, M.; and Choo, J. 2021. Viton-hd: High-resolution virtual try-on via misalignment-aware normalization. In Computer Vision and Pattern Recognition, 14131--14140

Show all 63 references
  1. [9]

    Choi, Y.; Kwak, S.; Lee, K.; Choi, H.; and Shin, J. 2024. Improving diffusion models for authentic virtual try-on in the wild. In European Conference on Computer Vision, 206--235

  2. [10]

    Chong, Z.; Dong, X.; Li, H.; Zhang, S.; Zhang, W.; Zhang, X.; Zhao, H.; Jiang, D.; and Liang, X. 2025. Catvton: Concatenation is all you need for virtual try-on with diffusion models. In International Conference on Learning Representations

  3. [11]

    Clancey, W. J. 1979. Transfer of Rule-Based Expertise through a Tutorial Dialogue . Ph.D. diss., Dept.\ of Computer Science, Stanford Univ., Stanford, Calif

  4. [12]

    Clancey, W. J. 1983. Communication, Simulation, and Intelligent Agents: Implications of Personal Intelligent Machines for Medical Education . In Proceedings of the Eighth International Joint Conference on Artificial Intelligence (IJCAI-83) , 556--560. Menlo Park, Calif: IJCAI ...

  5. [13]

    Clancey, W. J. 1984. Classification Problem Solving . In Proceedings of the Fourth National Conference on Artificial Intelligence, 45--54. Menlo Park, Calif.: AAAI Press

  6. [14]

    Clancey, W. J. 2021. The Engineering of Qualitative Models . Forthcoming

  7. [15]

    Engelmore, R.; and Morgan, A., eds. 1986. Blackboard Systems. Reading, Mass.: Addison-Wesley

  8. [16]

    Fuoli, D.; Van Gool, L.; and Timofte, R. 2021. Fourier space losses for efficient perceptual image super-resolution. In International Conference on Computer Vision, 2360--2369

  9. [17]

    Ge, C.; Song, Y.; Ge, Y.; Yang, H.; Liu, W.; and Luo, P. 2021 a . Disentangled cycle consistency for highly-realistic virtual try-on. In Computer Vision and Pattern Recognition, 16928--16937

  10. [18]

    Ge, Y.; Song, Y.; Zhang, R.; Ge, C.; Liu, W.; and Luo, P. 2021 b . Parser-free virtual try-on via distilling appearance flows. In Computer Vision and Pattern Recognition, 8485--8493

  11. [19]

    Gou, J.; Sun, S.; Zhang, J.; Si, J.; Qian, C.; and Zhang, L. 2023. Taming the power of diffusion models for high-quality virtual try-on with appearance flow. In ACM International Conference on Multimedia, 7599--7607

  12. [20]

    Gulrajani, I.; Ahmed, F.; Arjovsky, M.; Dumoulin, V.; and Courville, A. C. 2017. Improved training of wasserstein gans. Advances in Neural Information Processing Systems, 30

  13. [21]

    Han, X.; Wu, Z.; Wu, Z.; Yu, R.; and Davis, L. S. 2018. Viton: An image-based virtual try-on network. In Computer Vision and Pattern Recognition, 7543--7552

  14. [22]

    W.; Clancey, W

    Hasling, D. W.; Clancey, W. J.; and Rennels, G. 1984. Strategic explanations for a diagnostic consultation system. International Journal of Man-Machine Studies, 20(1): 3--19

  15. [23]

    W.; Clancey, W

    Hasling, D. W.; Clancey, W. J.; Rennels, G. R.; and Test, T. 1983. Strategic Explanations in Consultation---Duplicate . The International Journal of Man-Machine Studies, 20(1): 3--19

  16. [24]

    G.; and Wolff, G

    Hassibi, B.; Stork, D. G.; and Wolff, G. J. 1993. Optimal brain surgeon and general network pruning. In IEEE International Conference on Neural Networks, 293--299

  17. [25]

    He, S.; Song, Y.-Z.; and Xiang, T. 2022. Style-based global appearance flow for virtual try-on. In Computer Vision and Pattern Recognition, 3470--3479

  18. [26]

    Ho, J.; Jain, A.; and Abbeel, P. 2020. Denoising diffusion probabilistic models. Advances in Neural Information Processing Systems, 33: 6840--6851

  19. [27]

    Ho, J.; and Salimans, T. 2022. Classifier-free diffusion guidance. arXiv

  20. [28]

    Kim, J.; Gu, G.; Park, M.; Park, S.; and Choo, J. 2024. Stableviton: Learning semantic correspondence with latent diffusion model for virtual try-on. In Computer Vision and Pattern Recognition, 8176--8185

  21. [29]

    Kynk \"a \"a nniemi, T.; Aittala, M.; Karras, T.; Laine, S.; Aila, T.; and Lehtinen, J. 2024. Applying guidance in a limited interval improves sample and distribution quality in diffusion models. Advances in Neural Information Processing Systems, 37: 122458--122483

  22. [30]

    Lee, S.; Gu, G.; Park, S.; Choi, S.; and Choo, J. 2022. High-resolution virtual try-on with misalignment and occlusion-handled conditions. In European Conference on Computer Vision, 204--219

  23. [31]

    M.; Varadharajan, S.; and Kemelmacher-Shlizerman, I

    Lewis, K. M.; Varadharajan, S.; and Kemelmacher-Shlizerman, I. 2021. Tryongan: Body-aware try-on via layered interpolation. ACM Transactions on Graphics, 40(4): 1--10

  24. [32]

    Li, D.; Li, J.; and Hoi, S. 2023. Blip-diffusion: Pre-trained subject representation for controllable text-to-image generation and editing. Advances in Neural Information Processing Systems, 36: 30146--30166

  25. [33]

    Li, P.; Yan, S.; Tsai, J.; Zhang, R.; An, R.; Guo, Z.; and Gao, X. 2025. Adaptive Classifier-Free Guidance via Dynamic Low-Confidence Masking. arXiv preprint arXiv:2505.20199

  26. [34]

    Ma, R.; Zhang, Y.; Zhang, B.; Fang, L.; Huang, D.; and Qi, L. 2024. Learning attention in the frequency domain for flexible real photograph denoising. IEEE Transactions on Image Processing, 33: 3707--3721

  27. [35]

    Malarz, D.; Kasymov, A.; Zi e ba, M.; Tabor, J.; and Spurek, P. 2025. Classifier-free Guidance with Adaptive Scaling. arXiv preprint arXiv:2502.10574

  28. [36]

    Miyato, T.; Kataoka, T.; Koyama, M.; and Yoshida, Y. 2018. Spectral normalization for generative adversarial networks. arXiv preprint arXiv:1802.05957

  29. [37]

    Morelli, D.; Baldrati, A.; Cartella, G.; Cornia, M.; Bertini, M.; and Cucchiara, R. 2023. Ladi-vton: Latent diffusion textual-inversion enhanced virtual try-on. In ACM International Conference on Multimedia, 8580--8589

  30. [38]

    Morelli, D.; Fincato, M.; Cornia, M.; Landi, F.; Cesari, F.; and Cucchiara, R. 2022. Dress code: High-resolution multi-category virtual try-on. In Computer Vision and Pattern Recognition, 2231--2235

  31. [39]

    NASA . 2015. Pluto: The 'Other' Red Planet. https://www.nasa.gov/nh/pluto-the-other-red-planet. Accessed: 2018-12-06

  32. [40]

    W.; Hallacy, C.; Ramesh, A.; Goh, G.; Agarwal, S.; Sastry, G.; Askell, A.; Mishkin, P.; Clark, J.; et al

    Radford, A.; Kim, J. W.; Hallacy, C.; Ramesh, A.; Goh, G.; Agarwal, S.; Sastry, G.; Askell, A.; Mishkin, P.; Clark, J.; et al. 2021. Learning transferable visual models from natural language supervision. In International Conference on Machine Learning, 8748--8763

  33. [41]

    Ramesh, A.; Dhariwal, P.; Nichol, A.; Chu, C.; and Chen, M. 2022. Hierarchical text-conditional image generation with clip latents. arXiv, 1(2): 3

  34. [42]

    Rice, J. 1986. Poligon: A System for Parallel Problem Solving . Technical Report KSL-86-19, Dept.\ of Computer Science, Stanford Univ

  35. [43]

    Robinson, A. L. 1980 a . New Ways to Make Microcircuits Smaller. Science, 208(4447): 1019--1022

  36. [44]

    Robinson, A. L. 1980 b . New Ways to Make Microcircuits Smaller---Duplicate Entry . Science, 208: 1019--1026

  37. [45]

    Rombach, R.; Blattmann, A.; Lorenz, D.; Esser, P.; and Ommer, B. 2022. High-resolution image synthesis with latent diffusion models. In Computer Vision and Pattern Recognition, 10684--10695

  38. [46]

    Ronneberger, O.; Fischer, P.; and Brox, T. 2015. U-net: Convolutional networks for biomedical image segmentation. In Medical Image Computing and Computer Assisted Intervention, 234--241

  39. [47]

    L.; Ghasemipour, K.; Gontijo Lopes, R.; Karagol Ayan, B.; Salimans, T.; et al

    Saharia, C.; Chan, W.; Saxena, S.; Li, L.; Whang, J.; Denton, E. L.; Ghasemipour, K.; Gontijo Lopes, R.; Karagol Ayan, B.; Salimans, T.; et al. 2022. Photorealistic text-to-image diffusion models with deep language understanding. Advances in Neural Information Processing Syste...

  40. [48]

    Seitzer, M. 2020. pytorch-fid: FID Score for PyTorch. In Computer Vision and Pattern Recognition

  41. [49]

    Shen, D.; Song, G.; Xue, Z.; Wang, F.-Y.; and Liu, Y. 2024. Rethinking the spatial inconsistency in classifier-free diffusion guidance. In Computer Vision and Pattern Recognition, 9370--9379

  42. [50]

    Strang, G. 1993. Wavelet transforms versus Fourier transforms. Bulletin of the American Mathematical Society, 28(2): 288--305

  43. [51]

    Sun, K.; Cao, J.; Wang, Q.; Tian, L.; Zhang, X.; Zhuo, L.; Zhang, B.; Bo, L.; Zhou, W.; Zhang, W.; et al. 2024. Outfitanyone: Ultra-high quality virtual try-on for any clothing and any person. arXiv

  44. [52]

    N.; Kaiser, L.; and Polosukhin, I

    Vaswani, A.; Shazeer, N.; Parmar, N.; Uszkoreit, J.; Jones, L.; Gomez, A. N.; Kaiser, L.; and Polosukhin, I. 2017. Attention Is All You Need. arXiv:1706.03762

  45. [53]

    C.; Sheikh, H

    Wang, Z.; Bovik, A. C.; Sheikh, H. R.; and Simoncelli, E. P. 2004. Image quality assessment: from error visibility to structural similarity. IEEE Transactions on Image Processing, 13(4): 600--612

  46. [54]

    Xie, Z.; Huang, Z.; Dong, X.; Zhao, F.; Dong, H.; Zhang, X.; Zhu, F.; and Liang, X. 2023. Gp-vton: Towards general purpose virtual try-on via collaborative local-flow global-parsing learning. In Computer Vision and Pattern Recognition, 23550--23559

  47. [55]

    Xu, Y.; Gu, T.; Chen, W.; and Chen, A. 2025. Ootdiffusion: Outfitting fusion based latent diffusion for controllable virtual try-on. In AAAI Conference on Artificial Intelligence, volume 39, 8996--9004

  48. [56]

    Yang, H.; Zhang, R.; Guo, X.; Liu, W.; Zuo, W.; and Luo, P. 2020. Towards photo-realistic virtual try-on by adaptively generating-preserving image content. In Computer Vision and Pattern Recognition, 7850--7859

  49. [57]

    Yang, X.; Ding, C.; Hong, Z.; Huang, J.; Tao, J.; and Xu, X. 2024. Texture-preserving diffusion models for high-fidelity virtual try-on. In Computer Vision and Pattern Recognition, 7017--7026

  50. [58]

    Ye, H.; Zhang, J.; Liu, S.; Han, X.; and Yang, W. 2023. Ip-adapter: Text compatible image prompt adapter for text-to-image diffusion models. arXiv preprint arXiv:2308.06721

  51. [59]

    Zhang, L.; Rao, A.; and Agrawala, M. 2023. Adding conditional control to text-to-image diffusion models. In International Conference on Computer Vision, 3836--3847

  52. [60]

    A.; Shechtman, E.; and Wang, O

    Zhang, R.; Isola, P.; Efros, A. A.; Shechtman, E.; and Wang, O. 2018. The unreasonable effectiveness of deep features as a perceptual metric. In Computer Vision and Pattern Recognition, 586--595

  53. [61]

    Zhou, M.; Huang, J.; Yan, K.; Yu, H.; Fu, X.; Liu, A.; Wei, X.; and Zhao, F. 2022. Spatial-frequency domain information integration for pan-sharpening. In European Conference on Computer Vision, 274--291

  54. [62]

    W.; Xie, T.; Cong, Y.; Li, H.; Xu, M.; P \'e rez-R \'u a, J.-M.; et al

    Zhou, Z.; Liu, S.; Han, X.; Liu, H.; Ng, K. W.; Xie, T.; Cong, Y.; Li, H.; Xu, M.; P \'e rez-R \'u a, J.-M.; et al. 2025. Learning flow fields in attention for controllable person image generation. In Computer Vision and Pattern Recognition, 2491--2501

  55. [63]

    Zhu, L.; Yang, D.; Zhu, T.; Reda, F.; Chan, W.; Saharia, C.; Norouzi, M.; and Kemelmacher-Shlizerman, I. 2023. Tryondiffusion: A tale of two unets. In Computer Vision and Pattern Recognition, 4606--4615

Pith tools

Reviewed August 5, 2026 · model on record in the stance chip above.