Pith. sign in

REVIEW 3 major objections 2 minor 48 references

ModaFlow achieves higher-fidelity virtual try-on by applying distinct guidance rules to visual garment embeddings versus textual descriptions inside a flow-matching model.

Reviewed by Pith at T0; open to challenge. T0 means a machine referee read the full paper against a public rubric. the ladder, T0–T4 →

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

T0 review · grok-4.3

2026-06-29 05:08 UTC pith:TSAQYEZQ

load-bearing objection ModaFlow layers modality separation, two flow regularizers, and stochastic masks onto a flow-matching base for try-on, but the big FID drops are not shown to come from the modality split. the 3 major comments →

arxiv 2606.27773 v1 pith:TSAQYEZQ submitted 2026-06-26 cs.CV

ModaFlow: Modality-Aware Flow Matching for High-Fidelity Virtual Try-On

Xiangyu Sai , Meysam Madadi , Sergio Escalera , Yong Xu This is my paper
classification cs.CV
keywords virtual try-onflow matchingmodality-aware guidanceimage prompt adapterclassifier-free guidanceFID evaluationmask manipulationgarment preservation
verification ladder T0 review T1 audit T2 compute T3 formal T4 reserved

The pith

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

The paper introduces a flow-matching framework that keeps garment appearance and text descriptions aligned while the model adapts clothing to varied body shapes and large deformations. It does this by giving visual embeddings from a pretrained adapter fixed structural control and routing text embeddings through classifier-free guidance with adaptive scaling and zero-initialized velocity. Two new regularization losses enforce directional consistency and perceptual quality in the learned velocity field, and a stochastic mask schedule trains the model to handle different occlusion patterns at inference time. The resulting system reports lower FID scores than prior methods on both paired and unpaired test sets.

Core claim

ModaFlow shows that a modality-aware guidance scheme inside flow matching, where visual garment embeddings supply deterministic and persistent structural guidance while textual embeddings are modulated by classifier-free guidance with adaptive scaling and zero-initialized velocity, together with cosine-similarity and perceptual-flow-discrimination losses and stochastic sampling among box, transparent, and relaxed masks, produces more accurate velocity fields and higher-fidelity virtual try-on outputs than uniform conditioning approaches.

What carries the argument

The modality-aware guidance scheme that assigns persistent structural control to visual garment embeddings and adaptive classifier-free guidance to textual embeddings within the flow-matching velocity prediction.

Load-bearing premise

That visual garment embeddings extracted by a pretrained image prompt adapter deliver deterministic structural guidance that is reliably superior to treating visual and textual conditions the same way.

What would settle it

A controlled ablation on the same paired and unpaired benchmarks in which a single uniform conditioning pathway reaches FID values within 5 percent of the reported ModaFlow numbers would falsify the necessity of the modality-aware split.

Watch this falsifier — get emailed when new claim-graph text bears on it.

If this is right

  • The model reaches state-of-the-art qualitative and quantitative performance on standard virtual try-on benchmarks.
  • FID drops by roughly 30 percent on paired data and 20 percent on unpaired data.
  • The stochastic mask schedule allows reliable inference when only a box mask is supplied at test time.
  • The regularization losses improve both directional accuracy and perceptual realism of the predicted flow field.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

  • The same separation of deterministic visual guidance from scaled textual guidance could be tested in other conditional image-generation settings that combine reference images with descriptive text.
  • The mask manipulation strategy might generalize to training regimes that must handle variable occlusion without paired ground truth.
  • If the pretrained adapter's embeddings are replaced by a different visual encoder, the performance gap between modality-aware and uniform conditioning should be re-measured to isolate the contribution of the guidance split.

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

3 major / 2 minor

Summary. The paper introduces ModaFlow, a flow-matching framework for image-based virtual try-on that uses a modality-aware guidance scheme: pretrained visual garment embeddings supply deterministic structural guidance while textual embeddings are handled via classifier-free guidance with adaptive scaling and zero-initialized velocity. It adds cosine similarity and perceptual flow discrimination regularization losses plus stochastic sampling among box/transparent/relaxed masks during training. Experiments claim state-of-the-art qualitative and quantitative results, including ~30% FID reduction on paired and ~20% on unpaired benchmarks.

Significance. If the reported FID gains and qualitative improvements hold under controlled evaluation, the work would advance virtual try-on by explicitly separating visual and textual conditioning rather than treating modalities uniformly, with potential impact on e-commerce and AR applications. The concrete mechanisms (modality-aware scaling, zero-init velocity, dual regularization losses, and stochastic mask strategy) are clearly described contributions that could be adopted more broadly.

major comments (3)
  1. [Experiments] Experiments section (likely §4): the central SOTA claim of 30%/20% FID reductions is not supported by any ablation that isolates the modality-aware guidance scheme from the cosine/perceptual regularization losses or the stochastic mask sampling. Without such controls, the headline performance delta cannot be attributed to the proposed distinction between visual and textual modalities.
  2. [§3.2] §3.2 (modality-aware guidance): the claim that visual garment embeddings from the pretrained image prompt adapter supply 'deterministic, persistent structural guidance' under large clothing-body deformations is load-bearing for the method but is not validated by targeted metrics (e.g., deformation-specific alignment error) or failure-case analysis; the paper provides no evidence that this guidance remains superior when deformations exceed those seen in the adapter's training data.
  3. [Table 1] Table 1 / quantitative results: the reported FID numbers lack explicit confirmation that all baselines were re-trained or evaluated under identical dataset splits, resolution, and unpaired/paired protocols; the abstract supplies no such verification, undermining direct comparison.
minor comments (2)
  1. [§3.2] Notation for the adaptive scaling factor and zero-initialized velocity in the CFG formulation should be defined with an equation number for clarity.
  2. [§3.3] The description of the perceptual flow discrimination loss would benefit from an explicit equation showing how the discriminator is applied to the velocity field.

Simulated Author's Rebuttal

3 responses · 0 unresolved

We thank the referee for the constructive feedback on ModaFlow. The comments highlight opportunities to strengthen the experimental validation of our modality-aware guidance scheme. We respond point by point below and commit to revisions that directly address the concerns raised.

read point-by-point responses

  1. Referee: [Experiments] Experiments section (likely §4): the central SOTA claim of 30%/20% FID reductions is not supported by any ablation that isolates the modality-aware guidance scheme from the cosine/perceptual regularization losses or the stochastic mask sampling. Without such controls, the headline performance delta cannot be attributed to the proposed distinction between visual and textual modalities.

    Authors: We agree that the absence of component-wise ablations limits the ability to attribute gains specifically to the modality-aware guidance. In the revised version we will add a dedicated ablation study that systematically removes or disables the modality-aware scheme (while retaining the regularization losses and mask sampling) and reports the resulting FID changes on both paired and unpaired benchmarks. revision: yes

  2. Referee: [§3.2] §3.2 (modality-aware guidance): the claim that visual garment embeddings from the pretrained image prompt adapter supply 'deterministic, persistent structural guidance' under large clothing-body deformations is load-bearing for the method but is not validated by targeted metrics (e.g., deformation-specific alignment error) or failure-case analysis; the paper provides no evidence that this guidance remains superior when deformations exceed those seen in the adapter's training data.

    Authors: While the overall results demonstrate improved handling of deformations, we acknowledge that targeted quantitative validation is missing. We will augment §3.2 and the experiments with (i) a deformation-stratified alignment metric computed on high-deformation subsets and (ii) a short failure-case analysis comparing visual versus text-only conditioning under extreme pose changes. revision: yes

  3. Referee: [Table 1] Table 1 / quantitative results: the reported FID numbers lack explicit confirmation that all baselines were re-trained or evaluated under identical dataset splits, resolution, and unpaired/paired protocols; the abstract supplies no such verification, undermining direct comparison.

    Authors: All reported numbers were obtained under the same dataset splits, resolution, and paired/unpaired protocols described in §4. To make this explicit we will expand the evaluation protocol subsection and add a clarifying sentence to the Table 1 caption stating that every baseline was re-evaluated (or taken from the original paper when the authors released identical settings) under these controlled conditions. revision: yes

Circularity Check

0 steps flagged

No significant circularity; derivation self-contained

full rationale

The paper describes ModaFlow as an extension of an existing flow-matching base, introducing modality-aware guidance (visual embeddings vs. CFG for text), cosine/perceptual regularization losses, and stochastic mask sampling. No equations, derivations, or claims in the provided text reduce the FID improvements or SOTA results to quantities defined by fitted parameters from the same data, self-definitional loops, or load-bearing self-citations. All components are presented as independent additions without collapsing back to inputs by construction.

Axiom & Free-Parameter Ledger

0 free parameters · 1 axioms · 0 invented entities

The work relies on a pretrained image prompt adapter and standard flow-matching assumptions from prior literature; no new free parameters, axioms, or invented entities are explicitly introduced in the abstract.

axioms (1)
  • standard math Standard assumptions of flow matching and classifier-free guidance hold for the modality-aware scheme.
    The framework is built on existing flow matching and CFG techniques without new proofs.

pith-pipeline@v0.9.1-grok · 5750 in / 1200 out tokens · 23739 ms · 2026-06-29T05:08:36.843159+00:00 · methodology

0 comments
read the original abstract

Image-based virtual try-on has emerged as a compelling task in e-commerce and augmented reality, yet existing methods struggle to simultaneously preserve fine garment semantics and adapt to diverse person body geometries under large clothing-body deformations. We present ModaFlow, a modality-aware flow-matching based framework for high-fidelity virtual try-on that achieves precise alignment between textual descriptions and garment appearance. Unlike prior methods that treat multimodal conditions uniformly, ModaFlow introduces a modality-aware guidance scheme: visual garment embeddings extracted by a pretrained image prompt adapter provide deterministic, persistent structural guidance, while textual embeddings generated from garment descriptions are controlled via classifier-free guidance (CFG) with adaptive scaling and zero-initialized velocity. To further enhance flow field accuracy, we propose two regularization losses, cosine similarity and perceptual flow discrimination, that jointly improve directional consistency and perceptual realism of the velocity field. Additionally, a mask manipulation strategy stochastically samples among box, transparent, and relaxed masks during training, simulating diverse occlusion scenarios and enabling robust inference under unpaired settings where only a box mask is available. Experiments show that ModaFlow achieves state-of-the-art results in both qualitative and quantitative evaluations, reducing FID by approximately 30% on paired and 20% on unpaired benchmarks.

Figures

Figures reproduced from arXiv: 2606.27773 by Meysam Madadi, Sergio Escalera, Xiangyu Sai, Yong Xu.

Figure 1
Figure 1. Figure 1: Qualitative comparisons on four challenging cases left to right : (1) text/logo, (2) patterns, (3) bottoms, (4) in-the-wild. Top: representative recent SOTAs show visible artifacts or misalignment; Bottom : ModaFlow produces more plausible results with better detail preservation. Abstract. Image-based virtual try-on has emerged as a compelling task in e-commerce and augmented reality, yet existing methods … view at source ↗
Figure 2
Figure 2. Figure 2: ModaFlow pipeline. The model takes a triptych input consisting of a gar￾ment image, OpenPose map, and garment-agnostic person image generated through our stochastic mask manipulation strategy. Visual garment embeddings ev from the image prompt adapter provide persistent structural conditioning, while textual seman￾tics et are encoded with classifier-free guidance. The two modalities are fused through compl… view at source ↗
Figure 3
Figure 3. Figure 3: Three mask types used during train￾ing to improve robustness under diverse oc￾clusion conditions. From left to right in each batch: triptych input, triptych ground￾truth, auxiliary mask and garment. For clearer visualization, the transparency level in (b) is set to 40% instead of 10% used in training. Mask Sampling Strategy. Dur￾ing training, for each sample trip￾tych [\mathbf {g}, \mathbf {p}, \mathbf {x}… view at source ↗
Figure 4
Figure 4. Figure 4: Qualitative comparison on the VITON-HD dataset. The first row shows paired results, while the second and third rows show unpaired cross-category try-on cases (e.g. short-sleeve to long-sleeve). ModaFlow produces photo-realistic results with consistent body alignment and fine garment details across all settings. 5.3 Comparison with State-of-the-Art Methods Quantitative Results. As shown in [PITH_FULL_IMAGE… view at source ↗
Figure 5
Figure 5. Figure 5: Qualitative comparison on the DressCode dataset. The top row shows paired and the bottom row unpaired set￾tings. ModaFlow maintains garment fidelity and pose consistency across diverse cate￾gories and lighting conditions, outperform￾ing previous approaches in overall realism. Qualitative Results [PITH_FULL_IMAGE:figures/full_fig_p012_5.png] view at source ↗
Figure 6
Figure 6. Figure 6: In-the-wild try-on results. ModaFlow synthesizes plausible try-on outputs with better garment fidelity and alignment [PITH_FULL_IMAGE:figures/full_fig_p014_6.png] view at source ↗

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Reference graph

Works this paper leans on

48 extracted references · 5 canonical work pages · 2 internal anchors

  1. [1]

    In: ICLR 2023 Conference (2023)

    Albergo, M., Vanden-Eijnden, E.: Building normalizing flows with stochastic in- terpolants. In: ICLR 2023 Conference (2023)

    2023

  2. [2]

    In: International Conference on Learning Representations (2018)

    Bińkowski, M., Sutherland, D.J., Arbel, M., Gretton, A.: Demystifying mmd gans. In: International Conference on Learning Representations (2018)

    2018

  3. [3]

    IEEE Transactions on Pattern Analysis and Machine Intelligence (2019)

    Cao, Z., Hidalgo Martinez, G., Simon, T., Wei, S., Sheikh, Y.A.: Openpose: Real- time multi-person 2d pose estimation using part affinity fields. IEEE Transactions on Pattern Analysis and Machine Intelligence (2019)

    2019

  4. [4]

    In: Proc

    Choi, S., Park, S., Lee, M., Choo, J.: Viton-hd: High-resolution virtual try-on via misalignment-aware normalization. In: Proc. of the IEEE conference on computer vision and pattern recognition (CVPR) (2021)

    2021

  5. [5]

    In: European Conference on Computer Vision

    Choi, Y., Kwak, S., Lee, K., Choi, H., Shin, J.: Improving diffusion models for authentic virtual try-on in the wild. In: European Conference on Computer Vision. pp. 206–235. Springer (2024)

    2024

  6. [6]

    Catvton: Concatenation is all you need for virtual try-on with diffusion models.arXiv preprint arXiv:2407.15886, 2024

    Chong, Z., Dong, X., Li, H., Zhang, S., Zhang, W., Zhang, X., Zhao, H., Jiang, D., Liang, X.: Catvton: Concatenation is all you need for virtual try-on with diffusion models. arXiv preprint arXiv:2407.15886 (2024)

    work page arXiv 2024

  7. [7]

    In: The Thirteenth International Con- ference on Learning Representations (2025)

    Chung, H., Kim, J., Park, G.Y., Nam, H., Ye, J.C.: CFG++: Manifold-constrained classifier free guidance for diffusion models. In: The Thirteenth International Con- ference on Learning Representations (2025)

    2025

  8. [8]

    Advances in neural information processing systems34, 8780–8794 (2021)

    Dhariwal, P., Nichol, A.: Diffusion models beat gans on image synthesis. Advances in neural information processing systems34, 8780–8794 (2021)

    2021

  9. [9]

    In: Forty-first international conference on machine learning (2024)

    Esser, P., Kulal, S., Blattmann, A., Entezari, R., Müller, J., Saini, H., Levi, Y., Lorenz, D., Sauer, A., Boesel, F., et al.: Scaling rectified flow transformers for high-resolution image synthesis. In: Forty-first international conference on machine learning (2024)

    2024

  10. [10]

    CFG-Zero*: Improved Classifier-Free Guidance for Flow Matching Models

    Fan, W., Zheng, A.Y., Yeh, R.A., Liu, Z.: Cfg-zero*: Improved classifier-free guid- ance for flow matching models. arXiv preprint arXiv:2503.18886 (2025)

    work page Pith review arXiv 2025

  11. [11]

    Advances in neural in- formation processing systems27(2014)

    Goodfellow, I.J., Pouget-Abadie, J., Mirza, M., Xu, B., Warde-Farley, D., Ozair, S., Courville, A., Bengio, Y.: Generative adversarial nets. Advances in neural in- formation processing systems27(2014)

    2014

  12. [12]

    In: Proceed- ings of the IEEE/CVF International Conference on Computer Vision

    Guo,H.,Zeng,B.,Song,Y.,Zhang,W.,Liu,J.,Zhang,C.:Any2anytryon:Leverag- ing adaptive position embeddings for versatile virtual clothing tasks. In: Proceed- ings of the IEEE/CVF International Conference on Computer Vision. pp. 19085– 19096 (2025)

    2025

  13. [13]

    In: Proceedings of the IEEE conference on computer vision and pattern recognition

    Han, X., Wu, Z., Wu, Z., Yu, R., Davis, L.S.: Viton: An image-based virtual try-on network. In: Proceedings of the IEEE conference on computer vision and pattern recognition. pp. 7543–7552 (2018)

    2018

  14. [14]

    Advances in neural information processing systems33, 6840–6851 (2020)

    Ho, J., Jain, A., Abbeel, P.: Denoising diffusion probabilistic models. Advances in neural information processing systems33, 6840–6851 (2020)

    2020

  15. [15]

    Classifier-Free Diffusion Guidance

    Ho, J., Salimans, T.: Classifier-free diffusion guidance. arXiv preprint arXiv:2207.12598 (2022)

    work page internal anchor Pith review Pith/arXiv arXiv 2022

  16. [16]

    In: International Con- ference on Learning Representations (2022),https://openreview.net/forum?id= nZeVKeeFYf9

    Hu, E.J., Shen, Y., Wallis, P., Allen-Zhu, Z., Li, Y., Wang, S., Wang, L., Chen, W.: LoRA: Low-rank adaptation of large language models. In: International Con- ference on Learning Representations (2022),https://openreview.net/forum?id= nZeVKeeFYf9

    2022

  17. [17]

    CoRR (2024) 16 X

    Jiang, B., Hu, X., Luo, D., He, Q., Xu, C., Peng, J., Zhang, J., Wang, C., Wu, Y., Fu, Y.: Fitdit: Advancing the authentic garment details for high-fidelity virtual try-on. CoRR (2024) 16 X. Sai et al

    2024

  18. [18]

    Advances in Neural Information Processing Systems37, 52996–53021 (2024)

    Karras, T., Aittala, M., Kynkäänniemi, T., Lehtinen, J., Aila, T., Laine, S.: Guid- ing a diffusion model with a bad version of itself. Advances in Neural Information Processing Systems37, 52996–53021 (2024)

    2024

  19. [19]

    In: Proceedings of the IEEE/CVF conference on computer vision and pattern recognition

    Kim, J., Gu, G., Park, M., Park, S., Choo, J.: Stableviton: Learning semantic correspondence with latent diffusion model for virtual try-on. In: Proceedings of the IEEE/CVF conference on computer vision and pattern recognition. pp. 8176– 8185 (2024)

    2024

  20. [20]

    Labs, B.F.: Flux.https://github.com/black-forest-labs/flux(2024)

    2024

  21. [21]

    ACM Transactions on Graphics (TOG) 40(4), 1–10 (2021)

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

    2021

  22. [22]

    In: Proceedings of the IEEE/CVF Conference on Com- puter Vision and Pattern Recognition

    Liang, Y., Hu, X., Jiang, B., Luo, D., Peng, X., Wu, K., Xu, C., Han, W., Jin, T., Wang, C., et al.: Vton-handfit: Virtual try-on for arbitrary hand pose guided by hand priors embedding. In: Proceedings of the IEEE/CVF Conference on Com- puter Vision and Pattern Recognition. pp. 22616–22626 (2025)

    2025

  23. [23]

    In: 11th International Conference on Learning Representations, ICLR 2023 (2023)

    Lipman, Y., Chen, R.T., Ben-Hamu, H., Nickel, M., Le, M.: Flow matching for gen- erative modeling. In: 11th International Conference on Learning Representations, ICLR 2023 (2023)

    2023

  24. [24]

    Advances in Neural Information Processing Systems37, 133879– 133907 (2024)

    Liu, Q., Zeng, Z., He, J., Yu, Q., Shen, X., Chen, L.C.: Alleviating distortion in image generation via multi-resolution diffusion models and time-dependent layer normalization. Advances in Neural Information Processing Systems37, 133879– 133907 (2024)

    2024

  25. [25]

    In: The Eleventh International Conference on Learning Representations (ICLR) (2023)

    Liu, X., Gong, C., Liu, Q.: Flow straight and fast: Learning to generate and transfer data with rectified flow. In: The Eleventh International Conference on Learning Representations (ICLR) (2023)

    2023

  26. [26]

    CoRR (2025)

    Luan, J., Li, G., Zhao, L., Xing, W.: Mc-vton: Minimal control virtual try-on diffusion transformer. CoRR (2025)

    2025

  27. [27]

    In: Proceedings of the 41st International Conference on Machine Learning

    Mishchenko, K., Defazio, A.: Prodigy: an expeditiously adaptive parameter-free learner. In: Proceedings of the 41st International Conference on Machine Learning. pp. 35779–35804 (2024)

    2024

  28. [28]

    In: Proceedings of the Eu- ropean Conference on Computer Vision (2022)

    Morelli, D., Fincato, M., Cornia, M., Landi, F., Cesari, F., Cucchiara, R.: Dress Code: High-Resolution Multi-Category Virtual Try-On. In: Proceedings of the Eu- ropean Conference on Computer Vision (2022)

    2022

  29. [29]

    In: Proceedings of the IEEE/CVF conference on computer vision and pattern recognition

    Parmar, G., Zhang, R., Zhu, J.Y.: On aliased resizing and surprising subtleties in gan evaluation. In: Proceedings of the IEEE/CVF conference on computer vision and pattern recognition. pp. 11410–11420 (2022)

    2022

  30. [30]

    Peebles,W.,Xie,S.:Scalablediffusionmodelswithtransformers.In:Proceedingsof the IEEE/CVF international conference on computer vision. pp. 4195–4205 (2023)

    2023

  31. [31]

    In: The Twelfth International Conference on Learning Representations (2024)

    Podell, D., English, Z., Lacey, K., Blattmann, A., Dockhorn, T., Müller, J., Penna, J., Rombach, R.: Sdxl: Improving latent diffusion models for high-resolution image synthesis. In: The Twelfth International Conference on Learning Representations (2024)

    2024

  32. [32]

    In: ICLR (2023)

    Poole, B., Jain, A., Barron, J.T., Mildenhall, B.: Dreamfusion: Text-to-3d using 2d diffusion. In: ICLR (2023)

    2023

  33. [33]

    ACM Transactions on Multimedia Computing, Communications and Applications20(4), 1–20 (2023)

    Ren, B., Tang, H., Meng, F., Runwei, D., Torr, P.H., Sebe, N.: Cloth interac- tive transformer for virtual try-on. ACM Transactions on Multimedia Computing, Communications and Applications20(4), 1–20 (2023)

    2023

  34. [34]

    In: Proceedings of the Forty-second International Conference on Machine Learning (ICML) (2025) ModaFlow 17

    Ren, S., Yu, Q., He, J., Shen, X., Yuille, A., Chen, L.C.: Flowar: Scale-wise autore- gressive image generation meets flow matching. In: Proceedings of the Forty-second International Conference on Machine Learning (ICML) (2025) ModaFlow 17

    2025

  35. [35]

    In: ACM SIGGRAPH 2022 conference proceedings

    Sauer, A., Schwarz, K., Geiger, A.: Stylegan-xl: Scaling stylegan to large diverse datasets. In: ACM SIGGRAPH 2022 conference proceedings. pp. 1–10 (2022)

    2022

  36. [36]

    Insert Anything: Image Insertion via In-Context Editing in DiT

    Song, W., Jiang, H., Yang, Z., Quan, R., Yang, Y.: Insert anything: Image insertion via in-context editing in dit. arXiv preprint arXiv:2504.15009 (2025)

    work page Pith review arXiv 2025

  37. [37]

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

    Tran, M., Clements, J., Manoharan, A.P., Nguyen, T., Le, N.: Dualfit: A two- stage virtual try-on via warping and synthesis. In: Proceedings of the IEEE/CVF International Conference on Computer Vision. pp. 2397–2407 (2025)

    2025

  38. [38]

    In: Proceedings of the European conference on computer vision (ECCV)

    Wang, B., Zheng, H., Liang, X., Chen, Y., Lin, L., Yang, M.: Toward characteristic- preserving image-based virtual try-on network. In: Proceedings of the European conference on computer vision (ECCV). pp. 589–604 (2018)

    2018

  39. [39]

    Transactions on Machine Learning Research Journal (2024)

    Wang, X., Dufour, N., Andreou, N., Cani, M.P., Abrevaya, V.F., Picard, D., Kalo- geiton, V.: Analysis of classifier-free guidance weight schedulers. Transactions on Machine Learning Research Journal (2024)

    2024

  40. [40]

    IEEE transactions on image processing 13(4), 600–612 (2004)

    Wang, Z., Bovik, A.C., Sheikh, H.R., Simoncelli, E.P.: Image quality assessment: from error visibility to structural similarity. IEEE transactions on image processing 13(4), 600–612 (2004)

    2004

  41. [41]

    In: Proceedings of the Computer Vision and Pattern Recognition Conference

    Xia, M., Xue, N., Shen, Y., Yi, R., Gong, T., Liu, Y.J.: Rectified diffusion guidance for conditional generation. In: Proceedings of the Computer Vision and Pattern Recognition Conference. pp. 13371–13380 (2025)

    2025

  42. [42]

    In: Proceedings of the AAAI Conference on Artificial Intelligence

    Xu, Y., Gu, T., Chen, W., Chen, A.: Ootdiffusion: Outfitting fusion based latent diffusion for controllable virtual try-on. In: Proceedings of the AAAI Conference on Artificial Intelligence. vol. 39, pp. 8996–9004 (2025)

    2025

  43. [43]

    Qwen3 Technical Report

    Yang, A., Li, A., Yang, B., Zhang, B., Hui, B., Zheng, B., Yu, B., Gao, C., Huang, C., Lv, C., et al.: Qwen3 technical report. arXiv preprint arXiv:2505.09388 (2025)

    work page internal anchor Pith review Pith/arXiv arXiv 2025

  44. [44]

    In: Proceedings of the IEEE/CVF conference on computer vision and pattern recognition

    Yang, H., Zhang, R., Guo, X., Liu, W., Zuo, W., Luo, P.: Towards photo-realistic virtual try-on by adaptively generating-preserving image content. In: Proceedings of the IEEE/CVF conference on computer vision and pattern recognition. pp. 7850–7859 (2020)

    2020

  45. [45]

    In: Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR)

    Yang, X., Ding, C., Hong, Z., Huang, J., Tao, J., Xu, X.: Texture-preserving dif- fusion models for high-fidelity virtual try-on. In: Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR). pp. 7017–7026 (June 2024)

    2024

  46. [46]

    In: Proceedings of the IEEE conference on computer vision and pattern recognition

    Zhang, R., Isola, P., Efros, A.A., Shechtman, E., Wang, O.: The unreasonable effectiveness of deep features as a perceptual metric. In: Proceedings of the IEEE conference on computer vision and pattern recognition. pp. 586–595 (2018)

    2018

  47. [47]

    In: Proceedings of the Computer Vision and Pattern Recognition Conference

    Zhang, X., Song, D., Zhan, P., Chang, T., Zeng, J., Chen, Q., Luo, W., Liu, A.A.: Boow-vton: Boosting in-the-wild virtual try-on via mask-free pseudo data training. In: Proceedings of the Computer Vision and Pattern Recognition Conference. pp. 26399–26408 (2025)

    2025

  48. [48]

    In: Proceedings of the 41st International Conference on Machine Learning

    Zheng, C., Lan, Y.: Characteristic guidance: non-linear correction for diffusion model at large guidance scale. In: Proceedings of the 41st International Conference on Machine Learning. pp. 61386–61412 (2024)

    2024