Pith. sign in

REVIEW 4 major objections 5 minor 58 references

VITON-DRR: Details Retention Virtual Try-on via Non-rigid Registration

T0 review · 4 major / 5 minor · reviewed 2026-08-07 · deepseek-v4-flash

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

Pith's one-line read VITON-DRR claims that replacing learned thin-plate-spline warping with non-rigid point registration plus moving-least-squares deformation makes virtual try-on smoother and better at preserving garment details, reporting the best LPIPS…

desk verdict Plausible non-rigid registration warping for try-on, but the 'better than SOTA' claim is not supported by the paper's own table and its unusual LPIPS reference. read the letter →

arxiv 2505.23439 v1 pith:D56NOZ74 submitted 2025-05-29 cs.CV

Ben Li , Minqi Li , Jie Ren , Kaibing Zhang This is my paper · ORCID
classification cs.CV
keywords virtualtry-onnon-rigidregistrationmovingleastsquaresimagedeformationdetailpreservationsemanticsegmentationsynthesisgarmentwarping
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 proposes a virtual try-on method, VITON-DRR, whose central idea is to deform the target garment by explicitly registering its edge points to the edge of the target body region, then warping the whole image with moving least squares. This replaces the two dominant warp strategies in prior try-on systems — learned thin-plate-spline parameters and learned appearance flows — with a non-rigid point-cloud registration step that is not trained per dataset. The authors argue this yields smoother, more accurate deformation and preserves fine garment details such as stripes, text, and logos, which TPS and flow methods blur or lose under large pose changes and self-occlusion. On the Zalando test set the method ranks first on LPIPS and second on IS and SSIM among the compared methods, and ablations show both the semantic-segmentation and deformation modules contribute to the gain.

What carries the argument

The Deformation Module (DM) is the load-bearing component. It replaces learned warping with non-rigid registration: edge points are sampled (Canny detection plus curvature-based decimation) from the clothing mask and the target region mask; a Gaussian-mixture / expectation-maximization registration computes soft correspondences by minimizing an assignment-least-squares energy; and moving least squares uses those edge correspondences as control points to deform the full garment image with a similarity transform. The other two modules support it: the Human Semantics Generation Module produces a semantic segmentation of the dressed person, and the Image Synthesis Module, a conditional-GAN U-Net, fuses the warped garment with the clothing-agnostic person image. The assumption that edge-point correspondence predicts interior fabric deformation is what makes the DM work.

What would settle it

Compare, on a test set of garments with printed logos or text, the warped garment produced by the edge-registration pipeline against a manually warped ground truth: if interior pattern fidelity (e.g., OCR accuracy of garment text or LPIPS measured on the garment region only) is not better than the TPS and appearance-flow baselines, the claim that edge registration preserves interior details is not supported.

Watch

Extended reading notes

Core claim

The paper's central claim is that garment warping for virtual try-on is better done by geometric registration than by learned deformation networks: extract salient edge keypoints from the clothing mask and from the predicted target-region mask, match the two point sets with a Gaussian-mixture / expectation-maximization non-rigid registration, and use the resulting correspondences as moving-least-squares control points to warp the garment image. The authors assert that this pipeline handles large pose differences and self-occlusion better than TPS or appearance-flow baselines, keeps the warped garment visually coherent, and retains details; it is integrated with a conditional-GAN segmentation module that predicts the deformed human semantic layout and a synthesis module that fuses the warped garment with the person image.

Load-bearing premise

The load-bearing premise is that matching only the outer edge points of the garment to the outer edge of the target region also correctly dictates how the fabric inside those edges stretches, so that interior patterns, text, and textures deform faithfully.

Editorial extensions

If this is right

  • Virtual try-on systems can shift from learning warp parameters to explicit geometric registration, making deformation interpretable and independent of training data.
  • The best LPIPS score among the compared methods indicates the try-on output is perceptually closer to the target garment, while the second-place SSIM score leaves room for structural fidelity.
  • Because the deformation module is not learned, it can be applied to garment-person pairs without additional training, and its failures should localize to boundary correspondence errors.
  • The reported lower stability on long-sleeved and complex styles indicates the method's advantage is clearest for sleeveless and short-sleeved garments with well-separated edges.

Reading between the lines

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

  • The paper does not test it, but measuring LPIPS or text-legibility on the warped garment region alone would separate deformation quality from synthesis quality.
  • The paper does not test it, but adding interior keypoints at pattern corners or logo boundaries would directly test whether edge-only registration limits interior fidelity.
  • The paper does not test it, but the same registration-plus-MLS pipeline could be applied to texture transfer, logo or tattoo mapping, or frame-to-frame video warping wherever explicit point correspondences are available.
  • The paper does not test it, but the trade-off between the method's LPIPS lead and Flow-Style's SSIM lead suggests a human preference study or a detail-specific metric would clarify which gain matters for practical try-on.
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

4 major / 5 minor

Summary. VITON-DRR proposes a three-module virtual try-on pipeline: a human semantic segmentation generation module, a deformation module that extracts edge keypoints from the clothing and target-region masks and aligns them with the non-rigid registration algorithm of Li et al. (2020), and an image synthesis module based on a CGAN. The paper claims smoother and more accurate warping and better garment detail retention than existing methods, supporting this with qualitative comparisons on the Zalando dataset and quantitative IS, SSIM, and LPIPS scores. The code is publicly available.

Significance. The main strength of the paper is a clear, reproducible modular design that replaces learned warping with an explicit non-rigid registration plus moving-least-squares warping step; the public code makes independent verification feasible. However, the empirical support for the headline claim is weak: the reported metrics are mixed, the LPIPS reference is non-standard, and no direct quantitative evaluation of the warped garment is provided. The idea is defensible, but as submitted the evidence does not support the abstract's 'better than state-of-the-art' assertion.

major comments (4)
  1. [§4.3, Table 1] The abstract and conclusion state that VITON-DRR 'performs better than state-of-the-art methods', but Table 1 shows the method is second on IS (1.393 vs 1.421 for Flow-Style) and second on SSIM (0.779 vs 0.812 for Flow-Style), and its LPIPS advantage over DAFlow is only 0.002. No error bars or significance tests are reported, so the first-place LPIPS claim is not statistically established. Please either provide statistically grounded comparisons or soften the claims to match the table.
  2. [§4.3, LPIPS definition] The LPIPS is computed between the virtual try-on result and the flat target garment, not between the result and a ground-truth image of the same person wearing the garment. This is not the standard protocol for try-on evaluation and can reward under-warping, because a garment that remains closer to its flat product shape will tend to have lower LPIPS to the target clothes. Since no metric is reported on the warped garment W itself, the central claim that the deformation is 'more accurate' is not directly tested.
  3. [§4.5, Table 3] The ablation study shows small differences (IS 1.393 vs 1.357/1.352, SSIM 0.779 vs 0.754, LPIPS 0.575 vs 0.595/0.600) without error bars or significance tests, so it is not established that the HSGM and DM modules individually contribute beyond the baseline. Please report per-trial variance or a paired test, and show quantitative results on the warped garment (e.g., alignment error, segmentation IoU) rather than relying only on qualitative figures.
  4. [§3.2, Eqs. (2)-(5)] The deformation module is central to the method, yet the only evidence for its benefit is qualitative (Figs. 9 and 10) and the ablation table above. The claim in §3.2 that edge keypoints matched by the GMM/EM registration provide correct correspondences for interior texture deformation is never measured directly. A direct warping evaluation (e.g., keypoint transfer accuracy or warped-garment IoU with the target region) would make this load-bearing assumption testable.
minor comments (5)
  1. [Abstract] The phrase 'Compared with traditional methods' is grammatically incomplete; please revise to 'In comparison with traditional methods' or similar.
  2. [§4.1.2] The statement 'with loss weight λ = 1' is ambiguous because Eq. (1) defines λ1, Eq. (2) defines λ2, and the text does not specify σ2, α, EM iterations, or the edge-point sampling thresholds. Please report all hyperparameters for reproducibility.
  3. [§4.2] The sentence 'Under the condition of second-order difference constraints, the deformation effect of our DM method is better' is unclear because the DM module does not use second-order difference constraints; please rephrase to avoid confusion.
  4. [References] Several references have truncated or corrupted author names, including 'Pons-Mo, 2019' and 'Da, 2018'; please fix the bibliography entries.
  5. [Figure 7] The caption lists the methods in an order that does not match the panel labels in the figure; please align the caption with the figure layout.

Circularity Check

0 steps flagged · score 0.0 of 10

No circularity: the core registration is an externally published algorithm and the empirical comparison uses held-out test data; the LPIPS reference choice is a metric concern, not circularity.

full rationale

The paper's derivation chain does not reduce to its own inputs. The Deformation Module relies on the non-rigid registration method of Li et al. (2020), which is prior work by the corresponding author, but it is an externally published, parameter-free point-set registration algorithm with stated assumptions independent of the virtual try-on task. The paper does not invoke any self-cited uniqueness theorem to forbid alternatives, and it does not define the registration in terms of the try-on output. The warping uses Moving Least Squares from Schaefer et al. (2006). The quantitative comparison is performed on the held-out test split of the Zalando dataset, so the IS, SSIM, and LPIPS numbers are predictions on unseen data rather than fits. The paper's LPIPS measurement between the synthesized result and the flat target garment (Section 4.3) is a non-standard evaluation choice that may reward under-warping, but that is a correctness/benchmark concern, not a circular derivation. No load-bearing step is equivalent by construction to its own inputs, and no fitted parameter is renamed as a prediction. Therefore the appropriate circularity score is 0.

Assumptions & free parameters 6 free parameters · 6 assumptions · 0 invented entities

All constitutive parts are prior art: GMM/EM registration (Li et al. 2020), MLS (Schaefer et al. 2006), CGAN segmentation, and U-Net synthesis. The paper introduces no new free-standing entity. The main unverified assumptions are the sufficiency of edge points for texture warping and the chosen evaluation protocol.

free parameters (6)
  • lambda_1
    Weight balancing LCGAN and LCE in Eq. (1); value not reported.
  • lambda_2
    Regularization strength in Eq. (2) for the non-rigid registration energy; not reported.
  • sigma_squared
    Gaussian variance in the GMM matching likelihood, Eq. (3); controls soft assignment and is not reported.
  • alpha
    Deformation tolerance exponent in MLS weights, Eq. (5); authors only state alpha <= 1.
  • edge point sampling density/curvature threshold
    The paper says points with large curvature are sampled but does not specify the threshold or density.
  • EM iterations
    Number of iterations for the registration optimization is not given.
assumptions (6)
  • domain assumption The edge contours of the clothing mask and the target area mask are sufficient to determine the garment's deformation.
    Section 3.2 uses only edge points from Cm and GC for registration; interior texture deformation is assumed to follow.
  • domain assumption The GMM/EM registration algorithm from Li et al. (2020) yields correct point correspondences on garment contours even under large deformation and self-occlusion.
    Section 3.2 'Matching step' relies on this algorithm without modification or re-validation in this domain.
  • standard math MLS similarity transformations produce natural garment warping without local artifacts.
    Section 3.2 'Warping step' adopts MLS from Schaefer et al. (2006) as the deformation model.
  • ad hoc to paper LPIPS between the synthesized image and the target clothes measures garment detail retention.
    Section 4.3 defines the LPIPS reference as target clothes; this is not an established protocol.
  • domain assumption The predicted human semantic segmentation Sr from HSGM is accurate enough for warping.
    Section 3.1 assumes the generated Sr marks the target region correctly; the authors show qualitative examples only.
  • domain assumption The Zalando dataset annotations (pose map, segmentation) are correctly generated.
    Section 4.1.1 uses DensePose and pre-computed segmentation without checking for errors.

how reviews work

0 comments
Cite this review

Pith. "Pith review of VITON-DRR: Details Retention Virtual Try-on via Non-rigid Registration." pith.science (2026). https://pith.science/paper/D56NOZ74

@misc{pith2026250523439,
  author       = {Pith},
  title        = {Pith review of: VITON-DRR: Details Retention Virtual Try-on via Non-rigid Registration},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/D56NOZ74}},
  note         = {Machine review of arXiv:2505.23439}
}
read the original abstract

Image-based virtual try-on aims to fit a target garment to a specific person image and has attracted extensive research attention because of its huge application potential in the e-commerce and fashion industries. To generate high-quality try-on results, accurately warping the clothing item to fit the human body plays a significant role, as slight misalignment may lead to unrealistic artifacts in the fitting image. Most existing methods warp the clothing by feature matching and thin-plate spline (TPS). However, it often fails to preserve clothing details due to self-occlusion, severe misalignment between poses, etc. To address these challenges, this paper proposes a detail retention virtual try-on method via accurate non-rigid registration (VITON-DRR) for diverse human poses. Specifically, we reconstruct a human semantic segmentation using a dual-pyramid-structured feature extractor. Then, a novel Deformation Module is designed for extracting the cloth key points and warping them through an accurate non-rigid registration algorithm. Finally, the Image Synthesis Module is designed to synthesize the deformed garment image and generate the human pose information adaptively. {Compared with} traditional methods, the proposed VITON-DRR can make the deformation of fitting images more accurate and retain more garment details. The experimental results demonstrate that the proposed method performs better than state-of-the-art methods.

Figures

Figures reproduced from arXiv: 2505.23439 by the authors.

Figure 1
Figure 1. Example of missing clothing details in virtual try-on results. The fitting results [PITH_FULL_IMAGE:figures/full_fig_p004_1.png] view at source ↗
Figure 2
Figure 2. Examples of virtual try-on results synthesized by our method. Given a reference [PITH_FULL_IMAGE:figures/full_fig_p005_2.png] view at source ↗
Figure 3
Figure 3. An overview of our VITON-DRR, containing three main modules. 1. (a) [PITH_FULL_IMAGE:figures/full_fig_p009_3.png] view at source ↗
Figures from the paper (10 more)
Figure 4
Figure 4. Figure 4: The trajectories of the points are recorded and used to calculate the [PITH_FULL_IMAGE:figures/full_fig_p011_4.png]
Figure 4
Figure 4. Figure 4: Example of the clothes image registration. Given the clothes mask [PITH_FULL_IMAGE:figures/full_fig_p012_4.png]
Figure 5
Figure 5. Figure 5: Schematic diagram of the generator and discriminator of Image Synthesis Mod [PITH_FULL_IMAGE:figures/full_fig_p014_5.png]
Figure 6
Figure 6. Figure 6: Visualization of selected samples in the Zalando dataset. [PITH_FULL_IMAGE:figures/full_fig_p015_6.png]
Figure 7
Figure 7. Figure 7: Visual comparisons of CP-VITON, CP-VITON+, ACGPN, Flow-Style, DAFlow, [PITH_FULL_IMAGE:figures/full_fig_p017_7.png]
Figure 8
Figure 8. Figure 8: Comparison between our method and existing methods in virtual try-on for [PITH_FULL_IMAGE:figures/full_fig_p018_8.png]
Figure 9
Figure 9. Figure 9: Comparison of warping effects between TPS Han et al. (2018), STN Yang et al. [PITH_FULL_IMAGE:figures/full_fig_p019_9.png]
Figure 10
Figure 10. Figure 10: An example of the non-rigid point cloud registration with self-occlusions by [PITH_FULL_IMAGE:figures/full_fig_p020_10.png]
Figure 11
Figure 11. Figure 11: Impact of the human semantic segmentation module in the ablation study. [PITH_FULL_IMAGE:figures/full_fig_p022_11.png]
Figure 12
Figure 12. Figure 12: Visual comparison of virtual try-on results with and without the HSGM and [PITH_FULL_IMAGE:figures/full_fig_p023_12.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

58 extracted references · 55 canonical work pages

  1. [1]

    VTON-IT: Virtual Try-On using Image Translation

    author Adhikari, S. , author Bhusal, B. , author Ghimire, P. , author Shrestha, A. , year 2023 . title Vton-it: Virtual try-on using image translation . journal arXiv preprint arXiv:2310.04558

    work page Pith review arXiv 2023

  2. [2]

    , author Zhou, H

    author Bai, S. , author Zhou, H. , author Li, Z. , author Zhou, C. , author Yang, H. , year 2022 . title Single stage virtual try-on via deformable attention flows , in: booktitle Proc. of the European Conference on Computer Vision (ECCV) , organization Springer . pp. pages 409--425

    work page 2022

  3. [3]

    , author Tiwari, G

    author Bhatnagar, B.L. , author Tiwari, G. , author Theobalt, C. , author Pons-Mo , year 2019 . title Multi-garment net: Learning to dress 3d people from images , in: booktitle IEEE International Conference on Computer Vision ( ICCV ) , organization IEEE . pp. pages 5420--5430

    work page 2019

  4. [4]

    , year 1989

    author Bookstein, F.L. , year 1989 . title Principal warps: Thin-plate splines and the decomposition of deformations . journal In IEEE Trans. on Pattern Analysis and Machine Intelligence (PAMI) volume 11 , pages 567--585

    work page 1989

  5. [5]

    , author Cores, D

    author Bosquet, B. , author Cores, D. , author Seidenari, L. , author Brea, V.M. , author Mucientes, M. , author Del Bimbo, A. , year 2023 . title A full data augmentation pipeline for small object detection based on generative adversarial networks . journal Pattern Recognition volume 133 , pages 108998

    work page 2023

  6. [6]

    , author Wiesner, T

    author Bouchard, C. , author Wiesner, T. , author Desch \^e nes, A. , author Bilodeau, A. , author Turcotte, B. , author Gagn \'e , C. , author Lavoie-Cardinal, F. , year 2023 . title Resolution enhancement with a task-assisted gan to guide optical nanoscopy image analysis and acquisition . journal Nature Machine Intelligence volume 5 , pages 830--844

    work page 2023

  7. [7]

    , author Sheffer, A

    author Brouet, R. , author Sheffer, A. , author Boissieux, L. , author Cani, M.P. , year 2012 . title Design preserving garment transfer . journal ACM Trans. on Graphics volume 31 , pages 1--11

    work page 2012

  8. [8]

    , author Huang, Y

    author Chen, H. , author Huang, Y. , author Huang, H. , author Ge, X. , author Shao, D. , year 2024 a. title Gaussianvton: 3d human virtual try-on via multi-stage gaussian splatting editing with image prompting . journal arXiv preprint arXiv:2405.07472

    arXiv 2024

Show all 58 references
  1. [9]

    , author Zhang, X

    author Chen, J. , author Zhang, X. , author Ma, L. , author Yang, B. , author Zhang, K. , year 2024 b. title Cs-viton: a realistic virtual try-on network based on clothing region alignment and spm . journal The Visual Computer , pages 1--15

    2024

  2. [10]

    , author Ren, J

    author Chen, T. , author Ren, J. , year 2024 . title Integrating gan and texture synthesis for enhanced road damage detection . journal IEEE Trans. on Intelligent Transportation Systems (T-ITS) volume 25 , pages 12361--12371

    2024

  3. [11]

    , author Vaidhehi, M

    author Cherian, A.K. , author Vaidhehi, M. , author Arshey, M. , author Briskilal, J. , author Simpson, S.V. , year 2024 . title Generative adversarial networks with stochastic gradient descent with momentum algorithm for video-based facial expression . journal International J...

    2024

  4. [12]

    , author Jain, R

    author Chopra, A. , author Jain, R. , author Hemani, M. , author Krishnamurthy, B. , year 2021 . title Zflow: Gated appearance flow-based virtual try-on with 3d priors , in: booktitle IEEE International Conference on Computer Vision ( ICCV ) , organization IEEE . pp. pages 5433--5442

    2021

  5. [13]

    , author Rangarajan, A

    author Chui, H. , author Rangarajan, A. , year 2000 . title A feature registration framework using mixture models , in: booktitle Proceedings IEEE Workshop on Mathematical Methods in Biomedical Image Analysis (MMBIA) , pp. pages 190--197

    2000

  6. [14]

    , author Rangarajan, A

    author Chui, H. , author Rangarajan, A. , year 2003 . title A new point matching algorithm for non-rigid registration . journal Computer Vision and Image Understanding volume 89 , pages 114--141

    2003

  7. [15]

    , author Foruzan, A.H

    author Delavari, M. , author Foruzan, A.H. , author Chen, Y.W. , year 2019 . title Accurate point correspondences using a modified coherent point drift algorithm . journal Biomedical Signal Processing and Control volume 52 , pages 429--444

    2019

  8. [16]

    , author Liang, X

    author Dong, H. , author Liang, X. , author Shen, X. , author Wang, B. , author Lai, H. , author Zhu, J. , author Hu, Z. , author Yin, J. , year 2019 . title Towards multi-pose guided virtual try-on network , in: booktitle IEEE International Conference on Computer Vision ( ICC...

    2019

  9. [17]

    , author Geng, G

    author Du, H. , author Geng, G. , author Li, K. , author He, Y. , year 2013 . title 3d skull registration based on registration points automatic correspondence , in: booktitle 2013 International Conference on Virtual Reality and Visualization (ICVRV) , organization IEEE . pp. ...

    2013

  10. [18]

    , author Ma, J

    author Fan, A. , author Ma, J. , author Tian, X. , author Mei, X. , author Lin, W. , year 2022 . title Coherent point drift revisited for non-rigid shape matching and registration , in: booktitle IEEE Conf. on Computer Vision and Pattern Recognition (CVPR) , organization IEEE ...

    2022

  11. [19]

    , author Song, Y

    author Ge, Y. , author Song, Y. , author Zhang, R. , author Ge, C. , author Liu, W. , author Luo, P. , year 2021 . title Parser-free virtual try-on via distilling appearance flows , in: booktitle IEEE Conf. on Computer Vision and Pattern Recognition (CVPR) , organization IEEE ...

    2021

  12. [20]

    , author Neverova, N

    author G \"u ler, R.A. , author Neverova, N. , author Kokkinos, I. , year 2018 . title Densepose: Dense human pose estimation in the wild , in: booktitle IEEE Conf. on Computer Vision and Pattern Recognition (CVPR) , organization IEEE . pp. pages 7297--7306

    2018

  13. [21]

    , author Hu, X

    author Han, X. , author Hu, X. , author Huang, W. , author Scott, M.R. , year 2019 . title Clothflow: A flow-based model for clothed person generation , in: booktitle IEEE International Conference on Computer Vision ( ICCV ) , organization IEEE . pp. pages 10471--10480

    2019

  14. [22]

    , author Wu, Z

    author Han, X. , author Wu, Z. , author Wu, Z. , author Yu, R. , author Da , year 2018 . title Viton: An image-based virtual try-on network , in: booktitle IEEE Conf. on Computer Vision and Pattern Recognition (CVPR) , organization IEEE . pp. pages 7543--7552

    2018

  15. [23]

    , author Song, Y.Z

    author He, S. , author Song, Y.Z. , author Xiang, T. , year 2022 . title Style-based global appearance flow for virtual try-on , in: booktitle IEEE Conf. on Computer Vision and Pattern Recognition (CVPR) , organization IEEE . pp. pages 3470--3479

    2022

  16. [24]

    , author Li, H

    author Huang, Z. , author Li, H. , author Xie, Z. , author Kampffmeyer, M. , author Liang, X. , et al., year 2022 . title Towards hard-pose virtual try-on via 3d-aware global correspondence learning , in: booktitle Advances in Neural Information Processing Systems (NIPS) , org...

    2022

  17. [25]

    , author Ikenaga, T

    author Ishikawa, S. , author Ikenaga, T. , year 2022 . title Image-based virtual try-on system with clothing extraction module that adapts to any posture . journal Computers & Graphics volume 106 , pages 161--173

    2022

  18. [26]

    , author Bergmann, U

    author Jetchev, N. , author Bergmann, U. , year 2017 . title The conditional analogy gan: Swapping fashion articles on people images , in: booktitle IEEE International Conference on Computer Vision Workshops (ICCVW) , organization IEEE . pp. pages 2287--2292

    2017

  19. [27]

    , author Vemuri, B.C

    author Jian, B. , author Vemuri, B.C. , year 2011 . title Robust point set registration using gaussian mixture models . journal IEEE Transactions on Pattern Analysis and Machine Intelligence volume 33 , pages 1633--1645

    2011

  20. [28]

    , author Ba, J

    author Kingma, D.P. , author Ba, J. , year 2014 . title Adam: A method for stochastic optimization . journal arXiv preprint arXiv:1412.6980

    2014 arXiv

  21. [29]

    , author Gu, G

    author Lee, S. , author Gu, G. , author Park, S. , author Choi, S. , author Choo, J. , year 2022 . title High-resolution virtual try-on with misalignment and occlusion-handled conditions , in: booktitle Proc. of the European Conference on Computer Vision (ECCV) , organization ...

    2022

  22. [30]

    , author Xu, R.Y

    author Li, M. , author Xu, R.Y. , author Xin, J. , author Zhang, K. , author Jing, J. , year 2020 . title Fast non-rigid points registration with cluster correspondences projection . journal Signal Processing volume 170 , pages 107425

    2020

  23. [31]

    , author Caesar, H

    author Lin, Y. , author Caesar, H. , year 2024 . title Icp-flow: Lidar scene flow estimation with icp , in: booktitle IEEE Conf. on Computer Vision and Pattern Recognition (CVPR) , organization IEEE . pp. pages 15501--15511

    2024

  24. [32]

    , author Zhang, H

    author Liu, Y. , author Zhang, H. , author Xue, Y. , author Xu, G. , author Gao, Z. , year 2016 . title Registration of depth maps based on irls-icp-tps , in: booktitle 2016 9th International Congress on Image and Signal Processing, BioMedical Engineering and Informatics (CISP...

    2016

  25. [33]

    , author Zhao, J

    author Ma, J. , author Zhao, J. , author Yuille, A.L. , year 2016 . title Non-rigid point set registration by preserving global and local structures . journal IEEE Transactions on Image Processing volume 25 , pages 53--64

    2016

  26. [34]

    , author Corona, E

    author Marin, R. , author Corona, E. , author Pons-Moll, G. , year 2024 . title Nicp: Neural icp for 3d human registration at scale , in: booktitle Proc. of the European Conference on Computer Vision (ECCV) , organization Springer . pp. pages 265--285

    2024

  27. [35]

    , author Tuan, T.T

    author Minar, M.R. , author Tuan, T.T. , author Ahn, H. , author Rosin, P. , author Lai, Y.K. , year 2020 . title Cp-vton+: Clothing shape and texture preserving image-based virtual try-on , in: booktitle IEEE Conf. on Computer Vision and Pattern Recognition Workshops (CVPRW) ...

    2020

  28. [36]

    , author Osindero, S

    author Mirza, M. , author Osindero, S. , year 2014 . title Conditional generative adversarial nets . journal arXiv preprint arXiv:1411.1784

    2014 arXiv

  29. [37]

    , author Song, X

    author Myronenko, A. , author Song, X. , year 2010 . title Point set registration: Coherent point drift . journal IEEE Transactions on Pattern Analysis and Machine Intelligence volume 32 , pages 2262--2275

    2010

  30. [38]

    , author Liao, Z

    author Patel, C. , author Liao, Z. , author Pons-Moll, G. , year 2020 . title Tailornet: Predicting clothing in 3d as a function of human pose, shape and garment style , in: booktitle IEEE Conf. on Computer Vision and Pattern Recognition (CVPR) , organization IEEE . pp. pages ...

    2020

  31. [39]

    , author Pujades, S

    author Pons-Moll, G. , author Pujades, S. , author Hu, S. , author Black, M.J. , year 2017 . title Clothcap: Seamless 4d clothing capture and retargeting . journal ACM Trans. on Graphics volume 36 , pages 1--15

    2017

  32. [40]

    , author Goodfellow, I

    author Salimans, T. , author Goodfellow, I. , author Zaremba, W. , author Cheung, V. , author Radford, A. , author Chen, X. , year 2016 . title Improved techniques for training gans , in: booktitle Advances in Neural Information Processing Systems (NIPS) , organization Springe...

    2016

  33. [41]

    , author McPhail, T

    author Schaefer, S. , author McPhail, T. , author Warren, J. , year 2006 . title Image deformation using moving least squares . journal ACM Trans. on Graphics volume 25 , pages 533--540

    2006

  34. [42]

    , author Cresswell, J

    author Stein, G. , author Cresswell, J. , author Hosseinzadeh, R. , author Sui, Y. , author Ross, B. , author Villecroze, V. , author Liu, Z. , author Caterini, A.L. , author Taylor, E. , author Loaiza-Ganem, G. , year 2024 . title Exposing flaws of generative model evaluation...

    2024

  35. [43]

    , author Imaizumi, M

    author Takida, Y. , author Imaizumi, M. , author Shibuya, T. , author Lai, C.H. , author Uesaka, T. , author Murata, N. , author Mitsufuji, Y. , year 2024 . title SAN : Inducing metrizability of GAN with discriminative normalized linear layer , in: booktitle The Twelfth Intern...

    2024

  36. [44]

    , author Liu, M.Y

    author Tulyakov, S. , author Liu, M.Y. , author Yang, X. , author Kautz, J. , year 2018 . title Mocogan: Decomposing motion and content for video generation , in: booktitle IEEE Conf. on Computer Vision and Pattern Recognition (CVPR) , organization IEEE . pp. pages 1526--1535

    2018

  37. [45]

    , author Larsson, M

    author Usman Akbar, M. , author Larsson, M. , author Blystad, I. , author Eklund, A. , year 2024 . title Brain tumor segmentation using synthetic mr images-a comparison of gans and diffusion models . journal Scientific Data volume 11 , pages 259--287

    2024

  38. [46]

    , author Zheng, H

    author Wang, B. , author Zheng, H. , author Liang, X. , author Chen, Y. , author Lin, L. , author Yang, M. , year 2018 a. title Toward characteristic-preserving image-based virtual try-on network , in: booktitle Proc. of the European Conference on Computer Vision (ECCV) , orga...

    2018

  39. [47]

    , author Zhong, G

    author Wang, H. , author Zhong, G. , author Sun, J. , author Chen, Y. , author Zhao, Y. , author Li, S. , author Wang, D. , year 2023 . title Simultaneous restoration and super-resolution gan for underwater image enhancement . journal Frontiers in Marine Science volume 10 , pa...

    2023

  40. [48]

    , author Liu, M.Y

    author Wang, T.C. , author Liu, M.Y. , author Zhu, J.Y. , author Tao, A. , author Kautz, J. , author Catanzaro, B. , year 2018 b. title High-resolution image synthesis and semantic manipulation with conditional gans , in: booktitle IEEE Conf. on Computer Vision and Pattern Rec...

    2018

  41. [49]

    , author Bovik, A.C

    author Wang, Z. , author Bovik, A.C. , author Sheikh, H.R. , author Simoncelli, E.P. , year 2004 . title Image quality assessment: from error visibility to structural similarity . journal IEEE Trans. on Image Processing (TIP) volume 13 , pages 600--612

    2004

  42. [50]

    , author She, Q

    author Wang, Z. , author She, Q. , author Ward, T.E. , year 2021 . title Generative adversarial networks in computer vision: A survey and taxonomy . journal ACM Computing Surveys (CSUR) volume 54 , pages 1--38

    2021

  43. [51]

    , author Zhang, R

    author Yang, H. , author Zhang, R. , author Guo, X. , author Liu, W. , author Zuo, W. , year 2020 . title Towards photo-realistic virtual try-on by adaptively generating-preserving image content , in: booktitle IEEE Conf. on Computer Vision and Pattern Recognition (CVPR) , org...

    2020

  44. [52]

    , author Ong, S.H

    author Yang, Y. , author Ong, S.H. , author Foong, K.W.C. , year 2015 . title A robust global and local mixture distance based non-rigid point set registration . journal Pattern Recognition volume 48 , pages 156--173

    2015

  45. [53]

    , author Wang, X

    author Yu, R. , author Wang, X. , author Xie, X. , year 2019 . title Vtnfp: An image-based virtual try-on network with body and clothing feature preservation , in: booktitle IEEE International Conference on Computer Vision ( ICCV ) , organization IEEE . pp. pages 10511--10520

    2019

  46. [54]

    , author Zhan, J

    author Zhang, P. , author Zhan, J. , author Sun, K. , author Zhang, J. , author Wei, M. , author Wang, K. , year 2024 . title Gic-flow: Appearance flow estimation via global information correlation for virtual try-on under large deformation . journal Computers & Graphics volum...

    2024

  47. [55]

    , author Isola, P

    author Zhang, R. , author Isola, P. , author Efros, A.A. , author Shechtman, E. , author Wang, O. , year 2018 . title The unreasonable effectiveness of deep features as a perceptual metric , in: booktitle IEEE Conf. on Computer Vision and Pattern Recognition (CVPR) , organizat...

    2018

  48. [56]

    , author Chen, K

    author Zhou, Y. , author Chen, K. , author Xiao, R. , author Huang, H. , year 2023 . title Neural texture synthesis with guided correspondence , in: booktitle IEEE Conf. on Computer Vision and Pattern Recognition (CVPR) , organization IEEE . pp. pages 18095--18104

    2023

  49. [57]

    , author Park, T

    author Zhu, J.Y. , author Park, T. , author Isola, P. , author Efros, A.A. , year 2017 . title Unpaired image-to-image translation using cycle-consistent adversarial networks , in: booktitle IEEE International Conference on Computer Vision (ICCV) , organization IEEE . pp. page...

    2017

  50. [58]

    , author Xu, Z

    author Zhu, Z. , author Xu, Z. , author You, A. , author Bai, X. , year 2020 . title Semantically multi-modal image synthesis , in: booktitle IEEE Conf. on Computer Vision and Pattern Recognition (CVPR) , organization IEEE . pp. pages 5467--5476

    2020

Pith tools

Reviewed August 7, 2026 · model on record in the stance chip above.