Pith. sign in

REVIEW 5 major objections 5 minor 1 cited by

MGT: Extending Virtual Try-Off to Multi-Garment Scenarios

T0 review · 5 major / 5 minor · reviewed 2026-08-16 · deepseek-v4-flash

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

Pith's one-line read MGT claims that a single diffusion model with class-specific embeddings can reconstruct upper-body, lower-body, and dress garments at quality comparable to per-category models, while generalizing across datasets.

desk verdict A modest, legitimate multi-garment VTOFF extension with solid DressCode evidence, but the VITON-HD state-of-the-art claim does not survive contact with the same-split comparison. read the letter →

arxiv 2504.13078 v2 pith:B6KL5PUH submitted 2025-04-17 cs.CV cs.AI

Riza Velioglu , Petra Bevandic , Robin Chan , Barbara Hammer This is my paper · ORCID
classification cs.CVcs.AI
keywords virtualtry-offmulti-garmentreconstructionlatentdiffusionSigLIPconditioningclass-specificembeddingsperson-to-persontry-onDressCodeVITON-HD
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 introduces Multi-Garment TryOffDiff (MGT), a diffusion model that turns a photo of a clothed person into a clean, catalog-style image of a single garment. The central claim is that one unified model can reconstruct upper-body garments, lower-body garments, and dresses with quality close to that of models trained separately on each category. The authors also claim that MGT, trained only on the DressCode dataset, produces competitive results on the VITON-HD dataset without retraining. Achieving this would matter because virtual try-off currently requires a separate model per garment type, and because it enables person-to-person virtual try-on by supplying standardized garments extracted from ordinary photos.

What carries the argument

The load-bearing mechanism is the combination of SigLIP token conditioning with class-conditioned timestep embeddings. SigLIP features from the reference photo are projected by a lightweight adapter, computing $C(I) = (\mathrm{LN} \circ \mathrm{Linear} \circ \mathrm{SigLIP})(I)$, and injected into the U-Net's cross-attention layers as keys and values, following the IP-Adapter design. A 1,280-dimensional learnable embedding for the garment class is added element-wise to the timestep embedding ($e_t + e_c$) before it modulates each residual block, so the same network can steer generation toward the requested garment type. The SigLIP encoder and VAE are frozen; only the adapter, the class embeddings, and the U-Net are finetuned.

What would settle it

Take the DressCode test set, split it into plain garments and garments with visible logos or complex patterns, and compare MGT's DISTS scores on each subset against a category-specific TryOffDiff model; if MGT degrades substantially more on the logo/pattern subset, the SigLIP conditioning is the bottleneck. A second decisive test would be freezing the class embedding and varying only the reference image across categories: if outputs remain correct without the class label, the class embedding is not doing the claimed work; if outputs become ambiguous without it, the embedding is necessary.

Watch

Extended reading notes

Core claim

MGT is claimed to be the first unified model for multi-garment virtual try-off. It replaces text conditioning in a Stable Diffusion v1.4 latent diffusion backbone with SigLIP image features injected into cross-attention, and it adds a learnable embedding per garment class ('upper body', 'lower body', 'dress') to the timestep embedding in every residual block. On DressCode, MGT's scores are close to those of category-specific TryOffDiff models across all three categories (e.g., DISTS 22.2 vs 21.6 for upper body), and on VITON-HD it achieves better FID (21.9 vs 25.1) than a baseline trained on that dataset. The paper therefore positions MGT as a single model that replaces several specialized ones without a significant quality loss.

Load-bearing premise

The central assumption is that a frozen SigLIP encoder, through a single linear adapter, captures enough garment shape, texture, and pattern detail that the model can reconstruct high-fidelity catalog images without warping, segmentation, pose, or text guidance.

Editorial extensions

If this is right

  • One unified model can replace three per-category VTOFF models, lowering training and deployment cost for e-commerce catalog generation.
  • VTOFF models trained on one dataset can be applied to another dataset without finetuning, as shown by MGT's transfer from DressCode to VITON-HD.
  • Pairing MGT with a VTON model yields person-to-person try-on that avoids direct texture and skin-tone leakage from the source person, because garment extraction and person rendering are separated.
  • The class-conditioning trick is a small change to the timestep embedding and can likely be extended to new garment categories or attributes as annotated data becomes available.
  • MGT can turn user-generated photos into standardized product images, making catalog creation faster than studio photography.

Reading between the lines

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

  • If SigLIP features are the real bottleneck, then swapping in a higher-resolution or spatially denser encoder (e.g., patch-level features rather than a 77-token sequence) should improve logo and fine-texture fidelity without changing the architecture's class-conditioning mechanism.
  • The element-wise addition of class embeddings to timestep embeddings is a generic conditioning channel; the same mechanism could encode color, fabric, or silhouette attributes, turning VTOFF into an attribute-controllable generator.
  • The success on VITON-HD suggests that paired catalog data may be less necessary than assumed: a model trained on one domain could generate training pairs for another, potentially bootstrapping VTOFF in domains where product shots are rare.
  • Layered clothing is the natural stress test: the paper explicitly excludes it, and the class labels cannot specify which layer to extract, so a two-garment reference image with a jacket over a shirt would likely require explicit layering or ordering cues.
Share X Bluesky LinkedIn Reddit HN

Signed reviews

No signed human review yet.

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

5 major / 5 minor

Summary. The paper proposes Multi-Garment TryOffDiff (MGT), a latent diffusion model for virtual try-off that reconstructs catalog-style garment images from photos of clothed people. The model extends the authors' prior TryOffDiff by replacing text conditioning with SigLIP image features processed by a lightweight adapter and by adding learned class embeddings for upper-body, lower-body, and dresses, combined with the timestep embedding via element-wise addition (Eq. 5). Experiments compare MGT with category-specific TryOffDiff variants on DressCode (Table 1), with published baselines on VITON-HD (Table 2), and as a garment estimator feeding OOTDiffusion for person-to-person virtual try-on (Table 3). The central claims are that a single unified model achieves performance comparable to per-category models on DressCode, that it generalizes across domains to VITON-HD, and that it supports practical p2p-VTON.

Significance. If the DressCode result holds, the paper fills a clear gap: existing open VTOFF models handle only upper-body garments, whereas MGT is the first unified multi-category model. The Table 1 design—same training split, same architecture, only the unified-vs-specialized factor varied—is the right way to measure the cost of unification, and the release of demo, code, and models is a reproducible contribution. The limitations section honestly acknowledges that texture and logo fidelity remain challenging and that only three categories are covered. However, the cross-dataset and state-of-the-art claims are currently supported by an uncontrolled test-set comparison, and the main 'comparable performance' conclusion lacks variance or significance information; the overall contribution is useful but the evidence is partially overstated.

major comments (5)
  1. [Section 4.1, Table 2] MGT is evaluated on a deduplicated VITON-HD test set of 1,990 pairs, while the baseline numbers in Table 2 are taken from the original papers, which used the standard VITON-HD test split. Because FID, KID, and FDCLIP are distribution-level metrics, the favorable FID (21.9 vs. 25.1) and FDCLIP (7.0 vs. 9.4) results can be artifacts of test-set composition rather than model quality. This comparison does not support the cross-dataset generalization claim as stated.
  2. [Table 2, Abstract] The abstract claims 'state-of-the-art VTOFF results on VITON-HD,' but the same table shows MGT is worse than TryOffDiff on DISTS (24.7 vs. 23.0), LPIPS (36.3 vs. 32.4), SSIM (78.1 vs. 79.5), and equal on KID (8.9 vs. 8.9). The claim should be revised to 'competitive' or substantiated by a same-split re-evaluation of the baselines.
  3. [Section 4.2, Fig. 4] Inference hyperparameters (guidance scale and number of steps) are analyzed and selected by measuring FID and DISTS on DressCode-test, the same test set from which all reported quantitative results are computed. This is a test-set selection procedure; a validation split or nested protocol is needed before the reported numbers can be taken as unbiased estimates of MGT's performance.
  4. [Section 3.2, Eq. (5)] The paper does not ablate the class-embedding mechanism that is the paper's main technical novelty. Without a multi-category baseline trained without the class conditioning (or an alternative conditioning), it is not established that the element-wise addition in Eq. (5), rather than simply training one model on all categories, is responsible for the multi-garment behavior.
  5. [Table 1] No standard errors, confidence intervals, or significance tests are reported, and the differences are small (e.g., DISTS 22.2 vs. 21.6 for upper body, SSIM 80.2 vs. 80.8). The claim of 'comparable performance across all categories' needs variance information across seeds or bootstrapping to be established.
minor comments (5)
  1. [Section 5, Limitations] The limitations paragraph concedes that fine-grained texture recovery and logo preservation remain challenging and that only three garment categories are supported; this qualification should be reflected in the abstract's strong reconstruction language.
  2. [Table 2] Several metric entries for TryOffAnyone are missing (MS-SSIM, CW-SSIM, FDCLIP); please state whether these were not reported in the original paper or are not applicable.
  3. [Section 4.1] The adapter description says 1,024 SigLIP tokens are reduced to n = 77 conditioning embeddings; please clarify how the reduction is performed (learned projection, pooling, or truncation) and why 77 tokens were chosen.
  4. [Section 4.1] For p2p-VTON, the authors say ground-truth are unavailable and they randomly pair garments across individuals; the exact random pairing and whether the same pairs were used for all methods should be specified for reproducibility.
  5. [Section 3.2] The notation in Eq. (3) uses c both as the index set variable and in E_c; this is harmless but should be cleaned for clarity.

Circularity Check

0 steps flagged · score 2.0 of 10

No circular derivation: the multi-garment claim is tested against retrained per-category baselines on held-out DressCode data; the VITON-HD comparison is a test-set mismatch, not a circular step, and the self-citations are architectural rather than load-bearing.

full rationale

The central claim of the paper is that a single class-conditioned diffusion model (MGT) reconstructs upper-body, lower-body, and dress garments at quality comparable to per-category models. This is evaluated in Table 1 by training three category-specific TryOffDiff baselines on the same DressCode training split and comparing them to MGT on the held-out DressCode test set. The class embeddings are learned on the training split and are not derived from the test outputs, so the comparison is a controlled experiment that could have failed. No equation equates the predicted garment to an input: Eq. 5 merely adds a learned class vector to the timestep embedding, and the target garment image enters only through the standard MSE denoising loss. The paper's own limitation statement concedes that 'fine-grained texture recovery and logo preservation also remain challenging,' further showing that the conditional architecture is not assumed to be perfect by construction. The main self-citation is the statement in Section 3.2 that 'This conditioning has already proven effective for VTOFF results, as demonstrated by TryOffDiff [32].' This is a design justification, but it is not load-bearing for the DressCode result, which is independently established in Table 1. The VITON-HD generalization claim is weakened by the fact that MGT is evaluated on a deduplicated 1,990-pair test set ('we preprocessed the dataset by removing duplicates and test set leaks from the training set, resulting in ... 1,990 for testing') while baseline values are quoted from original papers ('Baseline values from original papers'), including the authors' own TryOffDiff. That is an invalid comparison and a correctness risk, especially for distribution-level metrics like FID, but it is not a reduction by construction or a fitted-input-as-prediction. Therefore the derivation chain is not circular; the score of 2 reflects the presence of minor, non-load-bearing self-citation in the cross-dataset comparison rather than any equation-level circularity.

Assumptions & free parameters 3 free parameters · 4 assumptions · 0 invented entities

The central claim depends on standard supervised-learning assumptions (dataset quality, metric validity) plus two model-specific choices: the SigLIP adapter as the only garment information conduit, and the additive class embedding as the disambiguation mechanism. No new physical or conceptual entities are introduced.

free parameters (3)
  • Garment class embeddings (3 vectors of dimension 1280) = Learned during training; values not reported
    Each of the three class labels ('upper body', 'lower body', 'dress') gets a learnable vector added to the timestep embedding (Eq. 5). This is the paper's core new mechanism.
  • Guidance scale s = 1.5
    Used for all main results; Fig. 4 explores s on the DressCode test set, so the selected value may be test-set tuned.
  • Inference steps n = 20
    Euler scheduler step count used during inference; also explored on the test set in Fig. 4.
assumptions (4)
  • domain assumption The VITON-HD and DressCode datasets provide correctly paired person and garment images, with garment annotations covering the full garment in the reference image.
    The model learns Q(G|C) from these pairs; if the ground-truth standardization is noisy or the label does not isolate a single garment, the reconstruction target is ill-defined. The paper mentions preprocessing VITON-HD to remove duplicates and test leaks, implying awareness of label imperfections.
  • domain assumption SigLIP features, after a single linear adapter, are sufficient to encode garment-specific shape, texture, and pattern information for reconstruction.
    The whole conditioning path is Eq. (1-2); no segmentation, pose, or warp is used. The paper's limitation about lost logos and textures shows this is a real assumption rather than a theorem.
  • ad hoc to paper Adding a per-class embedding to the timestep embedding is sufficient to disambiguate which garment in a multi-garment reference image to reconstruct.
    This is the proposed mechanism in Eq. 5; its adequacy is an architectural assumption validated only empirically on the three DressCode categories.
  • domain assumption FID, KID, DISTS, LPIPS, and SSIM computed against ground-truth garment images are valid measures of VTOFF reconstruction quality.
    The paper selects DISTS as primary but reports many metrics; perceptual metrics are known to be imperfect proxies and no human study is included.

how reviews work

0 comments
Cite this review

Pith. "Pith review of MGT: Extending Virtual Try-Off to Multi-Garment Scenarios." pith.science (2026). https://pith.science/paper/B6KL5PUH

@misc{pith2026250413078,
  author       = {Pith},
  title        = {Pith review of: MGT: Extending Virtual Try-Off to Multi-Garment Scenarios},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/B6KL5PUH}},
  note         = {Machine review of arXiv:2504.13078}
}
read the original abstract

Computer vision is transforming fashion industry through Virtual Try-On (VTON) and Virtual Try-Off (VTOFF). VTON generates images of a person in a specified garment using a target photo and a standardized garment image, while a more challenging variant, Person-to-Person Virtual Try-On (p2p-VTON), uses a photo of another person wearing the garment. VTOFF, in contrast, extracts standardized garment images from photos of clothed individuals. We introduce Multi-Garment TryOffDiff (MGT), a diffusion-based VTOFF model capable of handling diverse garment types, including upper-body, lower-body, and dresses. MGT builds on a latent diffusion architecture with SigLIP-based image conditioning to capture garment characteristics such as shape, texture, and pattern. To address garment diversity, MGT incorporates class-specific embeddings, achieving state-of-the-art VTOFF results on VITON-HD and competitive performance on DressCode. When paired with VTON models, it further enhances p2p-VTON by reducing unwanted attribute transfer, such as skin tone, ensuring preservation of person-specific characteristics. Demo, code, and models are available at: https://rizavelioglu.github.io/tryoffdiff/

Figures

Figures reproduced from arXiv: 2504.13078 by the authors.

Figure 1
Figure 1. Virtual try-off results generated by Multi-Garment TryOffDiff (MGT). The first three columns demonstrate MGT’s ability to generate multi-garment images (e.g., upper- and lower-body garments) from a single reference image. The last two columns illustrate dress reconstruction results. MGT generates an image in under 3 seconds on a consumer GPU. Please zoom in for a clearer view of details. Abstract Computer vision is … view at source ↗
Figure 2
Figure 2. Overview of various fashion image generation pipelines. First row illustrates Virtual Try-On, which synthesizes an image of a person wearing a target garment given input im￾ages of person and garment. Second row depicts Virtual Try-Off, which generates an e-commerce-style image of a garment from a photo of a person wearing it. Third row shows Person-to-Person Virtual Try-On, where a garment is transferred from one p… view at source ↗
Figure 3
Figure 3. Overview of MGT. Given a reference image and a class label (e.g. ‘dresses’), the SigLIP [42] image encoder extracts features, which are subsequently processed by a learnable adapter and embedded into the cross-attention layers of the Denoising U-Net. A learnable class label embedding conditions the generation to support multi-garment reconstruction. The model, with trainable Adapter, Embedding, and U-Net components,… view at source ↗
Figures from the paper (4 more)
Figure 4
Figure 4. Figure 4: Impact of guidance scale (s) and inference steps (n) on DISTS and FID scores. Evaluated on DressCode-test with MGT using the Euler scheduler [20] [PITH_FULL_IMAGE:figures/full_fig_p006_4.png]
Figure 5
Figure 5. Figure 5: Qualitative comparison of garment-specific and Multi-Garment TryOffDiff. First row displays the reference image. Second row shows dataset-specific reconstructions produced by TryOffDiff model trained on single category: VITON-HD (cols 1-3), DressCode upper-body (cols 4…
Figure 6
Figure 6. Figure 6: Qualitative comparison on VITON-HD. Despite not being trained on VITON-HD, MGT achieves competitive re￾sults compared to baselines specifically trained on it: TryOffD￾iff (ToD) [32] and TryOffAnyone (ToA) [35]. 5. Conclusion We introduced Multi-Garment TryOffDiff (MGT)…
Figure 7
Figure 7. Figure 7: Qualitative comparison for Person-to-Person Virtual Try-On in single and multi-garment settings. Columns: (i) source model image (person to be dressed), (ii) output from CatVTON, using a person image with target garments as condition for direct p2p￾VTON, (iii) output f…

Discussion (0). Continue with ORCID to comment.

Forward citations

Cited by 1 Pith paper

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

  1. Layering Virtual Try-On

    cs.CV 2026-07 conditional novelty 7.0 of 10

    A two-stage diffusion pipeline and new benchmark let virtual try-on add, remove, or swap clothing layers while preserving inner layers, with SOTA results on the new LVTON benchmark and on VITON-HD/DressCode.

Reference graph

Works this paper leans on

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

  1. [1]

    Layer normalization

    Jimmy Lei Ba, Jamie Ryan Kiros, and Geoffrey E Hinton. Layer normalization. stat, 1050:21, 2016. 4

    work page 2016

  2. [2]

    Multimodal garment designer: Human-centric latent diffusion models for fashion image editing

    Alberto Baldrati, Davide Morelli, Giuseppe Cartella, Mar- cella Cornia, Marco Bertini, and Rita Cucchiara. Multimodal garment designer: Human-centric latent diffusion models for fashion image editing. In ICCV, 2023. 3

    2023

  3. [3]

    Imagen 3

    Jason Baldridge, Jakob Bauer, Mukul Bhutani, Nicole Brich- tova, Andrew Bunner, Kelvin Chan, et al. Imagen 3. arXiv,

    work page

  4. [4]

    Shape matching and object recognition using shape contexts

    Serge Belongie, Jitendra Malik, and Jan Puzicha. Shape matching and object recognition using shape contexts. IEEE TPAMI, 2002. 3

    work page 2002

  5. [5]

    Improving image generation with better captions

    James Betker, Gabriel Goh, Li Jing, Tim Brooks, Jianfeng Wang, Linjie Li, et al. Improving image generation with better captions. preprint, 2023. 3

    work page 2023

  6. [6]

    Demystifying mmd gans

    Mikołaj Bi ´nkowski, Danica J Sutherland, Michael Arbel, and Arthur Gretton. Demystifying mmd gans. In ICLR, 2018. 5

    work page 2018

  7. [7]

    Vis- conet: Bridging and harmonizing visual and textual condi- tioning for controlnet

    Soon Yau Cheong, Armin Mustafa, and Andrew Gilbert. Vis- conet: Bridging and harmonizing visual and textual condi- tioning for controlnet. In ECCVW, 2024. 3

    work page 2024

  8. [8]

    Viton-hd: High-resolution virtual try-on via misalignment-aware normalization

    Seunghwan Choi, Sunghyun Park, Minsoo Lee, and Jaegul Choo. Viton-hd: High-resolution virtual try-on via misalignment-aware normalization. In CVPR, 2021. 1, 5

    work page 2021

Show all 49 references
  1. [9]

    Improving diffusion models for vir- tual try-on

    Yisol Choi, Sangkyung Kwak, Kyungmin Lee, Hyungwon Choi, and Jinwoo Shin. Improving diffusion models for vir- tual try-on. arXiv, 2024. https://doi.org/np47. 3

    2024

  2. [10]

    Catvton: Concatenation is all you need for virtual try-on with diffusion models

    Zheng Chong, Xiao Dong, Haoxiang Li, Shiyue Zhang, Wenqing Zhang, Xujie Zhang, Hanqing Zhao, and Xiaodan Liang. Catvton: Concatenation is all you need for virtual try-on with diffusion models. In ICLR, 2025. 3, 6

    2025

  3. [11]

    Image quality assessment: Unifying structure and texture similarity

    Keyan Ding, Kede Ma, Shiqi Wang, and Eero P Simoncelli. Image quality assessment: Unifying structure and texture similarity. IEEE TPAMI, 2020. 5

    2020

  4. [12]

    Fw-gan: Flow-navigated warping gan for video virtual try-on

    Haoye Dong, Xiaodan Liang, Xiaohui Shen, Bowen Wu, Bing-Cheng Chen, and Jian Yin. Fw-gan: Flow-navigated warping gan for video virtual try-on. In ICCV, 2019. 3

    2019

  5. [13]

    Scaling rectified flow transformers for high-resolution image synthesis

    Patrick Esser, Sumith Kulal, Andreas Blattmann, Rahim En- tezari, Jonas M¨uller, Harry Saini, et al. Scaling rectified flow transformers for high-resolution image synthesis. In ICML,

  6. [14]

    Parser-free virtual try-on via distilling appearance flows

    Yuying Ge, Yibing Song, Ruimao Zhang, Chongjian Ge, Wei Liu, and Ping Luo. Parser-free virtual try-on via distilling appearance flows. In CVPR, 2021. 3

    2021

  7. [15]

    Viton: An image-based virtual try-on network

    Xintong Han, Zuxuan Wu, Zhe Wu, Ruichi Yu, and Larry S Davis. Viton: An image-based virtual try-on network. In CVPR, 2018. 3

    2018

  8. [16]

    Clothflow: A flow-based model for clothed person generation

    Xintong Han, Xiaojun Hu, Weilin Huang, and Matthew R Scott. Clothflow: A flow-based model for clothed person generation. In CVPR, 2019. 3

    2019

  9. [17]

    Gans trained by a two time-scale update rule converge to a local nash equilib- rium

    Martin Heusel, Hubert Ramsauer, Thomas Unterthiner, Bernhard Nessler, and Sepp Hochreiter. Gans trained by a two time-scale update rule converge to a local nash equilib- rium. In NeurIPS, 2017. 5

    2017

  10. [18]

    Denoising diffu- sion probabilistic models

    Jonathan Ho, Ajay Jain, and Pieter Abbeel. Denoising diffu- sion probabilistic models. In NeurIPS, 2020. 5

    2020

  11. [19]

    The conditional analogy gan: Swapping fashion articles on people images

    Nikolay Jetchev and Urs Bergmann. The conditional analogy gan: Swapping fashion articles on people images. InICCVW,

  12. [20]

    Elucidating the design space of diffusion-based generative models

    Tero Karras, Miika Aittala, Timo Aila, and Samuli Laine. Elucidating the design space of diffusion-based generative models. In NeurIPS, 2022. 5, 6

    2022

  13. [21]

    Decoupled weight decay regularization

    Ilya Loshchilov and Frank Hutter. Decoupled weight decay regularization. In ICLR, 2019. 5

    2019

  14. [22]

    Dress code: High- resolution multi-category virtual try-on

    Davide Morelli, Matteo Fincato, Marcella Cornia, Federico Landi, Fabio Cesari, and Rita Cucchiara. Dress code: High- resolution multi-category virtual try-on. In CVPR, 2022. 1, 5

    2022

  15. [23]

    T2i-adapter: Learning adapters to dig out more controllable ability for text-to-image diffusion models

    Chong Mou, Xintao Wang, Liangbin Xie, Yanze Wu, Jian Zhang, Zhongang Qi, and Ying Shan. T2i-adapter: Learning adapters to dig out more controllable ability for text-to-image diffusion models. In AAAI, 2024. 3

    2024

  16. [24]

    Improved denoising diffusion probabilistic models

    Alexander Quinn Nichol and Prafulla Dhariwal. Improved denoising diffusion probabilistic models. In ICML, 2021. 2, 5

    2021

  17. [25]

    Zero-shot image-to-image translation

    Gaurav Parmar, Krishna Kumar Singh, Richard Zhang, Yijun Li, Jingwan Lu, and Jun-Yan Zhu. Zero-shot image-to-image translation. In SIGGRAPH, 2023. 3

    2023

  18. [26]

    Learn- ing transferable visual models from natural language super- vision

    Alec Radford, Jong Wook Kim, Chris Hallacy, Aditya Ramesh, Gabriel Goh, Sandhini Agarwal, Girish Sastry, Amanda Askell, Pamela Mishkin, Jack Clark, et al. Learn- ing transferable visual models from natural language super- vision. In ICML, 2021. 4

    2021

  19. [27]

    High-resolution image syn- thesis with latent diffusion models

    Robin Rombach, Andreas Blattmann, Dominik Lorenz, Patrick Esser, and Bj¨orn Ommer. High-resolution image syn- thesis with latent diffusion models. In CVPR, 2022. 2, 3, 5

    2022

  20. [28]

    Palette: Image-to-image diffusion models

    Chitwan Saharia, William Chan, Huiwen Chang, Chris Lee, Jonathan Ho, Tim Salimans, et al. Palette: Image-to-image diffusion models. In SIGGRAPH, 2022. 3

    2022

  21. [29]

    Image super- resolution via iterative refinement

    Chitwan Saharia, Jonathan Ho, William Chan, Tim Sali- mans, David J Fleet, and Mohammad Norouzi. Image super- resolution via iterative refinement. IEEE TPAMI, 2022. 3

    2022

  22. [30]

    Igr: Improving diffusion model for garment restoration from person image

    Le Shen, Rong Huang, and Zhijie Wang. Igr: Improving diffusion model for garment restoration from person image. arXiv preprint arXiv:2412.11513, 2024. 3

    2024 arXiv

  23. [31]

    Johnson, and Mao H

    Brandon Van Der Heide, Benjamin K. Johnson, and Mao H. Vang. The effects of product photographs and reputation sys- tems on consumer behavior and product cost on ebay. Com- put. Hum. Behav., 2013. 2

    2013

  24. [32]

    Tryoffdiff: Virtual-try-off via high-fidelity gar- ment reconstruction using diffusion models

    Riza Velioglu, Petra Bevandic, Robin Chan, and Barbara Hammer. Tryoffdiff: Virtual-try-off via high-fidelity gar- ment reconstruction using diffusion models. arXiv, 2024. https://doi.org/nt3n. 1, 2, 3, 4, 5, 6, 7

    2024

  25. [33]

    Toward characteristic- preserving image-based virtual try-on network

    Bochao Wang, Huabin Zheng, Xiaodan Liang, Yimin Chen, Liang Lin, and Meng Yang. Toward characteristic- preserving image-based virtual try-on network. In ECCV,

  26. [34]

    Bovik, H.R

    Zhou Wang, A.C. Bovik, H.R. Sheikh, and E.P. Simoncelli. Image quality assessment: from error visibility to structural similarity. IEEE Trans. Image Process., 2004. 5

    2004

  27. [35]

    Tryoffany- one: Tiled cloth generation from a dressed person

    Ioannis Xarchakos and Theodoros Koukopoulos. Tryoffany- one: Tiled cloth generation from a dressed person. arXiv,

  28. [36]

    Creating the best first impres- sion: Designing online product photos to increase sales

    Huosong Xia, Xiaoting Pan, Yanjun Zhou, and Zuopeng Justin Zhang. Creating the best first impres- sion: Designing online product photos to increase sales. Decis. Support Syst., 2020. 2

    2020

  29. [37]

    3, 5, 6, 7

    https://doi.org/n9bc. 3, 5, 6, 7

  30. [38]

    Prompt-free diffusion: Taking” text” out of text-to-image diffusion models

    Xingqian Xu, Jiayi Guo, Zhangyang Wang, Gao Huang, Ir- fan Essa, and Humphrey Shi. Prompt-free diffusion: Taking” text” out of text-to-image diffusion models. In CVPR, 2024. 3

    2024

  31. [39]

    Towards scalable unpaired virtual try-on via patch-routed spatially- adaptive gan

    Zhenyu Xie, Zaiyu Huang, Fuwei Zhao, Haoye Dong, Michael Kampffmeyer, and Xiaodan Liang. Towards scalable unpaired virtual try-on via patch-routed spatially- adaptive gan. In NeurIPS, 2021. 2, 3

    2021

  32. [40]

    Ip- adapter: Text compatible image prompt adapter for text-to- image diffusion models

    Hu Ye, Jun Zhang, Sibo Liu, Xiao Han, and Wei Yang. Ip- adapter: Text compatible image prompt adapter for text-to- image diffusion models. arXiv, 2023. https://doi. org/np3v. 3, 4

    2023

  33. [41]

    Oot- diffusion: Outfitting fusion based latent diffusion for control- lable virtual try-on

    Yuhao Xu, Tao Gu, Weifeng Chen, and Chengcai Chen. Oot- diffusion: Outfitting fusion based latent diffusion for control- lable virtual try-on. In AAAI, 2025. 3, 6

    2025

  34. [42]

    Sigmoid loss for language image pre-training

    Xiaohua Zhai, Basil Mustafa, Alexander Kolesnikov, and Lucas Beyer. Sigmoid loss for language image pre-training. In ICCV, 2023. 2, 4

    2023

  35. [43]

    Ti- legan: category-oriented attention-based high-quality tiled clothes generation from dressed person

    Wei Zeng, Mingbo Zhao, Yuan Gao, and Zhao Zhang. Ti- legan: category-oriented attention-based high-quality tiled clothes generation from dressed person. Neural Comput. Appl., 2020. 2

    2020

  36. [44]

    The unreasonable effectiveness of deep features as a perceptual metric

    Richard Zhang, Phillip Isola, Alexei A Efros, Eli Shechtman, and Oliver Wang. The unreasonable effectiveness of deep features as a perceptual metric. In CVPR, 2018. 5

    2018

  37. [45]

    Adding conditional control to text-to-image diffusion models

    Lvmin Zhang, Anyi Rao, and Maneesh Agrawala. Adding conditional control to text-to-image diffusion models. In ICCV, 2023. 3

    2023

  38. [46]

    Armani: Part-level garment-text alignment for unified cross-modal fashion design

    Xujie Zhang, Yu Sha, Michael C Kampffmeyer, Zhenyu Xie, Zequn Jie, Chengwen Huang, Jianqing Peng, and Xiaodan Liang. Armani: Part-level garment-text alignment for unified cross-modal fashion design. In ACMM, 2022. 2

    2022

  39. [47]

    Garmen- taligner: Text-to-garment generation via retrieval-augmented multi-level corrections

    Shiyue Zhang, Zheng Chong, Xujie Zhang, Hanhui Li, Yuhao Cheng, Yiqiang Yan, and Xiaodan Liang. Garmen- taligner: Text-to-garment generation via retrieval-augmented multi-level corrections. In ECCV, 2024. 3

    2024

  40. [49]

    Diffcloth: Diffusion based gar- ment synthesis and manipulation via structural cross-modal semantic alignment

    Xujie Zhang, Binbin Yang, Michael C Kampffmeyer, Wen- qing Zhang, Shiyue Zhang, Guansong Lu, Liang Lin, Hang Xu, and Xiaodan Liang. Diffcloth: Diffusion based gar- ment synthesis and manipulation via structural cross-modal semantic alignment. In ICCV, 2023. 2

    2023

  41. [2024]

    https://doi.org/nqr4. 3

Pith tools

Reviewed August 16, 2026 · model on record in the stance chip above.