Pith. sign in

REVIEW 4 major objections 5 minor 63 references

Training-Free, Identity-Preserving Image Editing for Fashion Pose Alignment and Normalization

T0 review · 4 major / 5 minor · reviewed 2026-08-10 · deepseek-v4-flash

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

Pith's one-line read A training-free, zero-shot pipeline normalizes long-sleeve garment poses in under a minute while preserving identity, texture, and branding.

desk verdict Plausible engineering integration for a narrow fashion task, but the evaluation never measures the pose-normalization or identity-preservation it claims to deliver. read the letter →

arxiv 2501.13692 v2 pith:DL3ODWB5 submitted 2025-01-23 cs.CV cs.AIcs.SE

Potito Aghilar , Vito Walter Anelli , Michelantonio Trizio , Eugenio Di Sciascio , Tommaso Di Noia This is my paper · ORCID
classification cs.CVcs.AIcs.SE
keywords fashionimageeditingposenormalizationtraining-freepipelinezero-shotdiffusionmodelsidentitypreservationgarmentpartscompositionlogorestoration
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 claims that converting a still-life product photo of a long-sleeve garment into a standardized 'arms at 45 degrees' pose can be done in under a minute, with no training and no fine-tuning, by chaining off-the-shelf diffusion and vision models with a few hand-written geometry steps. If true, fashion retailers could automate catalog image normalization at scale: the authors report the pipeline deployed on more than 30,000 garments at a global fashion retailer. The core bet is that identity preservation comes from the composition: generate a coarse edited pose, splice the original torso back into the image, and then restore logos in a separate step. The paper evaluates the pipeline against five established editing methods and argues that only this pipeline actually changes the pose while keeping the garment recognizable.

What carries the argument

The load-bearing mechanism is the garment parts-composition stage, Algorithm 1 in the paper, a fixed geometric routine that separates sleeves from torso by intersecting the garment mask with horizontal lines at 55% and 65% of the image height, shifting the vertical cut lines inward by 12 pixels, and then stitching the source torso with the generated sleeves in pixel space. Around this stage sits a conditioned unsampling procedure that injects and removes noise in the latent space to move from closed to open sleeves, a source-target shape matching routine that computes a scaling factor from torso bands to align the generated mask, and a logo detection-suppression-injection workflow built on off-the-shelf detection and segmentation models. The parts-composition stage is what enforces identity; the diffusion stages are what actualize the new pose.

What would settle it

Run the pipeline on a set of long-sleeve garments with complex textures (e.g., high-contrast plaids, large graphics) and with non-standard sleeve cuts (raglan, batwing), then inspect the seam where regenerated sleeves meet the original torso; if white gaps or duplicated texture appear on a substantial fraction of these cases, the central claim of consistent identity preservation breaks for that input class.

Watch

Extended reading notes

Core claim

On the paper's own terms, the central discovery is that a zero-shot, training-free pipeline can perform non-rigid pose editing on long-sleeve garments while preserving the garment's identity, texture, and branding. The pipeline works by decomposing the edit: a conditioned diffusion stage generates a coarse normalized-pose garment; a shape-matching stage rescales the result to align with the original silhouette; a parts-composition stage re-composites the original torso with the generated sleeves; and a logo workflow detects, suppresses, and re-injects brand marks. Because the torso pixels come from the source photo rather than from generation, the output retains the specific fabric and pattern of the original, while the regenerated sleeves provide the new pose.

Load-bearing premise

The pipeline assumes the fixed geometry of the parts-composition stage—two horizontal anchors at 55% and 65% height plus a 12-pixel inward shift—reliably separates sleeves from torso on every long-sleeve garment, and that this segmentation failure produces the visible gaps and duplicated textures the authors report.

Editorial extensions

If this is right

  • A fashion brand can normalize entire product catalogs without curating paired training data, since the pipeline requires only a single still-life photo per garment.
  • The under-one-minute runtime on a single GPU makes the pipeline a practical batch preprocessing step for e-commerce workflows rather than a research demo.
  • Because the torso is reused from the original photo, high-frequency identity details such as fabric weave, prints, and stitching survive the edit in a way that pure latent-space edits do not.
  • The separate logo workflow decouples brand preservation from pose editing, allowing the same architecture to be reused for any object with a distinct 'identity region' to protect.

Reading between the lines

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

  • The fixed anchor lines at 0.55h and 0.65h assume a standard catalog silhouette; garments with raglan or batwing sleeves will likely fall outside this assumption, which is exactly the class of errors visible in the paper's own limitation figure.
  • The pipeline's design pattern—use a generative model to hypothesize the new pose, then use classical pixel-space composition to guarantee identity—could transfer to other non-rigid object edits, such as repositioning straps on bags or laces on shoes.
  • A testable extension is to replace the heuristic segmentation with a learning-based part segmenter; if that removes the reported white-gap and texture-dup artifacts, the pipeline's remaining bottleneck is the texture reconstruction of generated sleeves.
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. The paper proposes FashionRepose, a training-free, zero-shot multi-stage pipeline that normalizes the pose of long-sleeve garment images from a still-life configuration to a standardized 45-degree arm-torso angle while preserving garment identity, texture, and branding. The pipeline chains off-the-shelf models (Realistic Vision, ControlNet, IP-Adapter, Florence2, SAM2, 4xUltrasharp) with custom computer-vision modules for sleeve/torso segmentation, shape matching, parts composition, and logo detection/suppression/re-injection. The authors report deployment at OVS on over 30,000 garments and provide qualitative comparisons and an ablation study on DressCode and VITON-HD.

Significance. If validated, the work addresses a real industrial need: converting product photos to a standardized pose without per-domain training data, using only off-the-shelf components. The paper benefits from a clearly defined task, a detailed and reproducible-sounding pipeline, explicit discussion of limitations, and an ablation that shows each stage's contribution. However, the scientific evidence is currently insufficient: the quantitative evaluation does not measure pose accuracy or identity preservation, and the only task-level support is qualitative. The claim of being entirely training-free is also overstated because a classification stage is fine-tuned on DressCode. The central idea is plausible and the weaknesses appear addressable, but the paper needs substantially stronger evaluation and internal consistency before publication.

major comments (4)
  1. [Sec. 4.2 (Table 1) and Sec. 4.4 (Table 2)] The quantitative results for the full pipeline are inconsistent between the main evaluation and the ablation study. On DressCode upper-body, Table 1 reports Ours LPIPS 0.3147, PSNR 30.2503, SSIM 0.7212, whereas Table 2 reports the final pipeline (comprising coarse generation + conditional unsampling + parts composition + logo restoration) as LPIPS 0.1976, PSNR 33.4796, SSIM 0.8052. On VITON-HD the discrepancy is similarly large: LPIPS 0.3396 in Table 1 versus 0.2106 in Table 2. The paper does not explain this discrepancy (e.g., different subsets, random seeds, or evaluation protocols), and it is not a minor point because the ablation's claimed improvements only make sense if the baseline and final results are measured under the same protocol. The authors must clarify or reconcile these numbers.
  2. [Sec. 4.2 (Table 1) and Sec. 4.3] No task-level metric supports the central claim of identity-preserving pose normalization. LPIPS, PSNR, and SSIM are computed against the source image, not against a ground-truth target pose, and Sec. 4.2 acknowledges they are used 'without a ground truth reference.' These metrics therefore cannot verify that the output reaches the intended 45-degree arm-torso angle or that torso texture and logos are preserved. On these generic metrics, FashionRepose scores worse than nearly all baselines (e.g., DressCode LPIPS 0.3147 vs. 0.2735 for FPE; VITON-HD LPIPS 0.3396 vs. 0.2744 for TIC). The statement that baselines 'are unable to solve the pose normalization task' is asserted but not quantified. The paper needs either a task-specific evaluation (e.g., pose keypoint distance to a target template, garment segmentation IoU for the preserved torso, or a controlled user study) or an explicit explanation of why the reported generic metrics are not relevant.
  3. [Abstract, Sec. 3.3, and Supplemental Sec. 7.2] The claim that FashionRepose is 'training-free' and eliminates 'the requirement for specialized training data' is contradicted by the long-sleeve detection stage, which uses a VGG16 CNN 'trained on a subset of DressCode' and described in the supplement as fine-tuned (achieving 97% accuracy). The paper calls the filter 'not mandatory,' but it is part of the pipeline as described (Sec 3.2, Stage 1) and is needed for deployment in existing workflows. The authors should either remove this trained component from the core pipeline and clarify that detection is optional, or revise the claim to state that only the pose-editing stages are training-free.
  4. [Sec. 3.8 and Sec. 5] The parts-composition algorithm (Algorithm 1) is the principal mechanism for preserving garment identity, because the torso and logos are pasted from the original image. However, its robustness is not quantified: the fixed horizontal anchor lines at 0.55h and 0.65h, the 12-pixel padding, and the mask-intersection logic are all heuristics that, as the authors note in Sec. 5 and Fig. 10, can produce white gaps or duplicated sleeve textures when the mask misaligns. Since a failure here directly compromises the central claim of identity preservation, the paper should report the frequency and severity of such artifacts (e.g., mask alignment IoU statistics, or a failure-rate analysis over the test set) rather than relying on selected qualitative examples.
minor comments (5)
  1. [Sec. 3.6] The term 'unsampling' is unusual and not defined in the main text; please clarify that it refers to the process of adding conditioned noise to move the latents in a desired direction, and use the spelling consistently throughout.
  2. [Algorithm 1 (Sec. 3.8)] The condition in line 6, 'if r and t intercept 12 points on the mask,' is unclear: what exactly are the 12 points, and how is the intersection counted? Please specify the precise condition, as this determines when the fallback default mask is used.
  3. [Sec. 3.4] The input requirements (1:1 aspect ratio, white background, minimum 1024x1024 resolution) limit the applicability of the method, but they are presented as specifications without discussion of how restrictive they are for real-world catalogs; please add a sentence on expected failure modes for inputs outside these specifications.
  4. [Supplemental Sec. 7.2 (Fig. 12)] The prompt text contains a typo: 'Be coincise' should read 'Be concise.'
  5. [Sec. 3.2 and Sec. 4.1] The claim of processing 'in less than a minute' is not backed by a measured timing or complexity analysis; please report the actual runtime per image on the described RTX 4090 setup.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: FashionRepose is an engineering pipeline with no derivation that reduces to its inputs, no fitted parameter renamed as prediction, and no load-bearing self-citation chain.

full rationale

I walked the claimed derivation chain in arXiv:2501.13692. The paper makes no first-principles mathematical claim; it presents a multi-stage image-editing pipeline (Sec. 3). The core claim—that the pipeline performs training-free, zero-shot, identity-preserving pose normalization—is supported by an explicit construction: the torso region of the original still-life image is composited back into the output (Algorithm 1, Sec. 3.8), and logos are detected, suppressed, and re-injected (Sec. 3.10). Identity preservation in the torso is therefore achieved by direct pixel reuse, not by circular reasoning: the paper openly describes this as a composition step rather than disguising it as a prediction. No stage fits a parameter to a quantity that is later reported as an independent prediction. The quantitative evaluation (LPIPS/PSNR/SSIM, Table 1) compares outputs to the source image without a ground-truth target pose, and the baselines actually score better on these generic metrics; the authors rest their task-level claim on qualitative comparisons (Sec. 4.3, Figs. 8, 16–18). This is an evaluation-design weakness and a correctness risk, but it is not circularity: the pipeline's outputs are not defined as the metric values, and no claimed result is equivalent to an input by construction. I also checked the reference list for self-citation chains: the cited prior work (e.g., MasaCtrl, Null-Text Inversion, TIC, FPE, ControlNet, IP-Adapter, Florence2, SAM2) is external, and no load-bearing premise is justified only by a citation to the present authors' own prior work. The deployment sentence about OVS is unverifiable in the manuscript but is an assertion, not a circular derivation. Overall, the central claim has independent engineering content, and the limitations the authors acknowledge (mask misalignment, duplicate sleeves, wrong logo injection, Sec. 5, Fig. 10) further confirm that the pipeline is not tautologically successful. No circular step meets the evidentiary bar of quoting a specific reduction or fitted-input-renamed-as-prediction, so the circularity score is 0.

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

The pipeline's novel content is a composition of hand-tuned heuristics over pretrained models. Its free parameters are geometric constants in the shape-matching and parts-composition stages, and its core assumptions concern the reliability of off-the-shelf models and the input image constraints. No new entities, forces, or fitted physical laws are introduced.

free parameters (4)
  • Anchor line heights in Algorithm 1 = 0.55h and 0.65h
    Hand-chosen horizontal lines A-B and C-D in Algorithm 1 (Sec 3.8) define the sleeve/torso boundary; no sensitivity analysis is provided.
  • Padding factor P = 12 pixels
    Algorithm 1 shifts the vertical boundaries inward by 12 pixels (Sec 3.8); value is ad hoc and not justified.
  • Band width b in shape matching = 30 pixels
    The scaling factor in source-target shape matching uses a 30-pixel torso band (Sec 3.7); selection criterion not reported.
  • Conditioned unsampling timestep = not reported
    The unsampling stage (Sec 3.6) requires a noise-injection timestep; the value is not given, so the stage cannot be exactly reproduced.
assumptions (4)
  • domain assumption Off-the-shelf models (Realistic Vision, ControlNet, IP-Adapter Plus, SAM2, Florence2) are reliable on fashion still-life images with white backgrounds.
    The entire zero-shot pipeline depends on pretrained models generalizing to this specific input distribution (Sec 3.5, 3.6, 3.10).
  • domain assumption Input images satisfy resolution >= 1024x1024, 1:1 aspect ratio, and solid white background.
    These are stated requirements in Sec 3.4; violation degrades or breaks the pipeline.
  • domain assumption The standardized 45-degree arm-to-torso angle is a valid and sufficient target pose for 'normalization'.
    Task definition in Sec 3.1; the target is chosen for the industry use case and is not validated against user or designer preferences.
  • domain assumption Source garments are in a still-life pose with sleeves relaxed close to the body, leaving torso and sleeve regions separable.
    Task definition in Sec 3.1 and Algorithm 1 assume a canonical source pose; other poses would break the line-based segmentation.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Training-Free, Identity-Preserving Image Editing for Fashion Pose Alignment and Normalization." pith.science (2026). https://pith.science/paper/DL3ODWB5

@misc{pith2026250113692,
  author       = {Pith},
  title        = {Pith review of: Training-Free, Identity-Preserving Image Editing for Fashion Pose Alignment and Normalization},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/DL3ODWB5}},
  note         = {Machine review of arXiv:2501.13692}
}
read the original abstract

Diffusion models have recently unlocked new possibilities in editing images of real-world objects. Yet, transforming objects in non-rigid ways, such as modifying poses or applying image-based conditioning, continues to present significant challenges. Retaining the unique identity of objects during these edits is a complex task, and current techniques often fall short of delivering the precision needed for industrial settings, where consistency is non-negotiable. Additionally, adapting diffusion models demands custom training data, which is often unavailable in real-world scenarios. To address these gaps, we present FashionRepose, a novel, training-free pipeline designed to handle non-rigid pose adjustments specifically for the fashion industry. This approach combines pretrained off-the-shelf models to modify the poses of long-sleeve garments while safeguarding their identity and branding characteristics. By adopting a zero-shot methodology, FashionRepose enables near real-time edits, entirely eliminating the requirement for specialized training data. FashionRepose has been deployed for a global fashion firm, OVS, handling more than 30,000 long-sleeve garments.

Figures

Figures reproduced from arXiv: 2501.13692 by the authors.

Figure 1
Figure 1. Pose Normalization Task Overview. FASHIONREPOSE presents a training-free, consistency-aware method for normalizing the pose of long-sleeve garments, starting from an initial still-life configuration. Our pipeline enforces a standardized 45-degree arm-to-torso alignment, preserving garment identity, texture fidelity, and brand-specific attributes across complex, non-rigid transformations (Sec. 3.2). Combining state-o… view at source ↗
Figure 2
Figure 2. Proposed Pipeline Architecture. The figure depicts the architecture of FASHIONREPOSE, a pose-normalization pipeline tailored for fashion domain. The pipeline is able to edit, in less than a minute, the pose of long-sleeve garments maintaining consistency, and identity. The entire workflow is entirely based on pretrained off-the-shelf models fostering ease of adoption through its training-free capability. The stages … view at source ↗
Figure 3
Figure 3. Coarse Garment Generation. Leveraging Realistic Vision v5.1, we generate a first coarse pose-normalized clothing starting from a still-life logo-suppressed garment. During this stage, the Stable Diffusion model is conditioned by different em￾beddings coming from both the IP-Adapter and ControlNets. the torso in the initial still-life pose, the adoption of a con￾ditioned diffusion model was necessary to reconstruct t… view at source ↗
Figures from the paper (17 more)
Figure 5
Figure 5. Figure 5: Garment Parts-Composition. The figure illustrates a graphical representation of Algorithm 1. Left: The points iden￾tified by the algorithm are over-imposed on the normalized-pose mask. Right: The green lines represent the discriminating bound￾aries between the sleeves …
Figure 6
Figure 6. Figure 6: Source-Target Shape Matching. Left: Silhouette com￾parison showcasing the discrepancies between source and target garments prior to applying the shape matching algorithm. Right: Silhouette alignment after the shape matching algorithm, illustrat￾ing the successful trans…
Figure 4
Figure 4. Figure 4: Conditioned Unsampling. From the still-life and the pose-normalized images, we obtain the latents and the canny￾processed images. The latents are fused together through a gradient-mask blending operation. Then, a two-pass guiding pro￾cess (unsampling - sampling) adds a…
Figure 7
Figure 7. Figure 7: ). Logo Detection. Detecting the logo in the original still-life images is essential to preserve brand identity across edits. To accomplish such a task, an approach based on Florence2 and SAM2 is proposed. Florence2 detects the logo in the 5https://huggingface.co/Kim20…
Figure 8
Figure 8. Figure 8: Qualitative Evaluation. The figure depicts the obtained results from a qualitative perspective on different baselines. Top: a garment sampled from DressCode [17] (upper-body only) dataset. Bottom: a garment sampled from VITON-HD [6]. Our approach is the only one capabl…
Figure 9
Figure 9. Figure 9: Qualitative Ablation Study. The figure depicts the obtained results from a qualitative perspective through an ablation study of the pipeline stages. Top: a garment sampled from DressCode [17] (upper-body only) dataset. Bottom: a garment sampled from VITON￾HD [6]. From …
Figure 10
Figure 10. Figure 10: Limitations of the Pipeline. The artifacts are high￾lighted in red. Left: The output presents wrong duplicated sleeves with a different texture pattern, due to a highly complex shape and texture of the original garment. Right: The neck is detected as a logo, injecting…
Figure 11
Figure 11. Figure 11: Prompt for Baselines. The figure showcases the struc￾tured text prompt used to perform pose edits with various text-to￾image and pose-to-image baselines, i.e., MasaCtrl [3], Null-Text Inversion [30], TIC [8], FPE [27], and ControlNet [52]. 7.2. Long Sleeve Detection T…
Figure 12
Figure 12. Figure 12: Prompt for Long Sleeve Detection. This figure presents the text prompt used in the second step of long sleeve detection, combining a preamble and a detailed caption generated by Florence2. The prompt is used by LLaMa3-8b to determine whether the garment in the image c…
Figure 13
Figure 13. Figure 13: Prompt for Logo Detection. The figure illustrates the text prompt designed to guide Florence2 in the logo detection phase of the FASHIONREPOSE pipeline. The prompt lists items Florence2 should attempt to identify within a still-life garment image. These items include:…
Figure 15
Figure 15. Figure 15: Prompt for Conditional Unsampling. The figure de￾tails the positive and negative text prompts used to guide the con￾ditional unsampling stage. This stage refines the garment’s pose transformation in the latent space to ensure visual fidelity and con￾sistency with the …
Figure 16
Figure 16. Figure 16: Qualitative Evaluation on DressCode Dataset. The figure illustrates the results, for the baselines, from a qualitative perspec￾tive. The process is executed on DressCode [17] (upper-body only) dataset. The utilized prompts are discussed in Sec. 7.1. (a) Reference (b) …
Figure 17
Figure 17. Figure 17: Qualitative Evaluation on VITON-HD Dataset. The figure illustrates the results, for the baselines, from a qualitative perspective. The process is executed on VITON-HD [6]. The utilized prompts are discussed in Sec. 7.1. 4 [PITH_FULL_IMAGE:figures/full_fig_p015_17.png]
Figure 18
Figure 18. Figure 18: Qualitative Evaluation on VITON-HD Dataset with Logos. The figure illustrates the results, for the baselines, from a qualitative perspective. The process is executed on a logo-only subset of VITON-HD [6]. The utilized prompts are discussed in Sec. 7.1. (a) Reference (…
Figure 19
Figure 19. Figure 19: Qualitative Ablation on DressCode Upper-Body Dataset. The figure illustrates the results of the ablation study, for different stages of the pipeline, from a qualitative perspective. The process is executed on DressCode [17] (upper-body only) dataset. 5 [PITH_FULL_IMA…
Figure 20
Figure 20. Figure 20: Qualitative Ablation on VITON-HD Full Dataset. The figure illustrates the results of the ablation study, for different stages of the pipeline, from a qualitative perspective. The process is executed on the full VITON-HD [6] dataset. (a) Reference (b) Coarse generation…
Figure 21
Figure 21. Figure 21: Qualitative Ablation on VITON-HD Dataset with Logos. The figure illustrates the results of the ablation study, for different stages of the pipeline, from a qualitative perspective. The process is executed on a logo-only subset of the VITON-HD [6] dataset. 6 [PITH_FUL…

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

63 extracted references · 52 canonical work pages

  1. [1]

    Gimp 2.8 for photographers: Image editing with open source software

    Marie Altenburg. Gimp 2.8 for photographers: Image editing with open source software. PSA Journal, 80(1):9–10, 2014. 2

    work page 2014

  2. [2]

    Steerable Conditional Diffusion for Out-of-Distribution Adaptation in Medical Image Reconstruction

    Riccardo Barbano, Alexander Denker, Hyungjin Chung, Tae-Hoon Roh, Simon Arrdige, Peter Maass, Bangti Jin, and Jong Chul Ye. Steerable conditional diffusion for out-of- distribution adaptation in imaging inverse problems. CoRR, abs/2308.14409, 2023. 2

    work page Pith review arXiv 2023

  3. [3]

    Masactrl: Tuning-free mu- tual self-attention control for consistent image synthesis and editing

    Mingdeng Cao, Xintao Wang, Zhongang Qi, Ying Shan, Xi- aohu Qie, and Yinqiang Zheng. Masactrl: Tuning-free mu- tual self-attention control for consistent image synthesis and editing. In IEEE/CVF International Conference on Com- puter Vision, ICCV 2023, Paris, France, October 1-6, 2023, pages 22503–22513. IEEE, 2023. 3, 7, 1

    work page 2023

  4. [4]

    Image editing with adobe photoshop 6.0

    Ronald D Caruso and Gregory C Postel. Image editing with adobe photoshop 6.0. Radiographics, 22(4):993–1002,

    work page

  5. [5]

    Reproducible scal- ing laws for contrastive language-image learning

    Mehdi Cherti, Romain Beaumont, Ross Wightman, Mitchell Wortsman, Gabriel Ilharco, Cade Gordon, Christoph Schuh- mann, Ludwig Schmidt, and Jenia Jitsev. Reproducible scal- ing laws for contrastive language-image learning. In Pro- ceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pages 2818–2829, 2023. 6

    work page 2023

  6. [6]

    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 Proc. of the IEEE conference on computer vision and pattern recognition (CVPR), 2021. 7, 8, 3, 4, 5, 6

    work page 2021

  7. [7]

    Diffu- sion models beat gans on image synthesis

    Prafulla Dhariwal and Alexander Quinn Nichol. Diffu- sion models beat gans on image synthesis. In Advances in Neural Information Processing Systems 34: Annual Con- ference on Neural Information Processing Systems 2021, NeurIPS 2021, December 6-14, 2021, virtual , pages 8780– 8794, 2021. 2

    work page 2021

  8. [8]

    Tuning-free inversion-enhanced control for consistent image editing

    Xiaoyue Duan, Shuhao Cui, Guoliang Kang, Baochang Zhang, Zhengcong Fei, Mingyuan Fan, and Junshi Huang. Tuning-free inversion-enhanced control for consistent image editing. In Thirty-Eighth AAAI Conference on Artificial In- telligence, AAAI 2024, Thirty-Sixth Conference on Innova- tive Applications of Artificial Intelligence, IAAI 2024, Four- teenth Sympo...

    work page 2024

Show all 63 references
  1. [9]

    Abhimanyu Dubey, Abhinav Jauhri, Abhinav Pandey, Ab- hishek Kadian, Ahmad Al-Dahle, Aiesha Letman, Akhil Mathur, Alan Schelten, Amy Yang, Angela Fan, Anirudh Goyal, Anthony Hartshorn, Aobo Yang, Archi Mitra, Archie Sravankumar, Artem Korenev, Arthur Hinsvark, Arun Rao, Aston Z...

    2024 arXiv

  2. [10]

    Scaling rec- tified flow transformers for high-resolution image synthesis

    Patrick Esser, Sumith Kulal, Andreas Blattmann, Rahim Entezari, Jonas M ¨uller, Harry Saini, Yam Levi, Dominik Lorenz, Axel Sauer, Frederic Boesel, Dustin Podell, Tim Dockhorn, Zion English, and Robin Rombach. Scaling rec- tified flow transformers for high-resolution image syn...

    2024

  3. [11]

    C-vton: Context-driven image-based virtual try-on network

    Benjamin Fele, Ajda Lampe, Peter Peer, and Vitomir Struc. C-vton: Context-driven image-based virtual try-on network. In Proceedings of the IEEE/CVF Winter Conference on Ap- plications of Computer Vision (WACV) , pages 3144–3153,

  4. [12]

    Disentangled cycle consistency for highly- realistic virtual try-on

    Chongjian Ge, Yibing Song, Yuying Ge, Han Yang, Wei Liu, and Ping Luo. Disentangled cycle consistency for highly- realistic virtual try-on. In 2021 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) , pages 16923–16932, 2021

    2021

  5. [13]

    Parser-free virtual try-on via distilling ap- pearance flows

    Yuying Ge, Yibing Song, Ruimao Zhang, Chongjian Ge, Wei Liu, and Ping Luo. Parser-free virtual try-on via distilling ap- pearance flows. In2021 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) , pages 8481–8489,

  6. [14]

    Goodfellow, Jean Pouget-Abadie, Mehdi Mirza, Bing Xu, David Warde-Farley, Sherjil Ozair, Aaron C

    Ian J. Goodfellow, Jean Pouget-Abadie, Mehdi Mirza, Bing Xu, David Warde-Farley, Sherjil Ozair, Aaron C. Courville, and Yoshua Bengio. Generative adversarial networks.CoRR, abs/1406.2661, 2014. 2, 3 9

    2014 arXiv

  7. [15]

    Ai assisted fashion design: A re- view

    Ziyue Guo, Zongyang Zhu, Yizhi Li, Shidong Cao, Hangyue Chen, and Gaoang Wang. Ai assisted fashion design: A re- view. IEEE Access, 11:88403–88415, 2023. 2

    2023

  8. [16]

    Xintong Han, Zuxuan Wu, Zhe Wu, Ruichi Yu, and Larry S. Davis. Viton: An image-based virtual try-on network. In 2018 IEEE/CVF Conference on Computer Vision and Pat- tern Recognition, pages 7543–7552, 2018. 2

    2018

  9. [17]

    Dresscode: Autoregressively sewing and gener- ating garments from text guidance

    Kai He, Kaixin Yao, Qixuan Zhang, Jingyi Yu, Lingjie Liu, and Lan Xu. Dresscode: Autoregressively sewing and gener- ating garments from text guidance. ACM Trans. Graph., 43 (4):72:1–72:13, 2024. 3, 7, 8, 4, 5

    2024

  10. [18]

    Imagine yourself: Tuning-Free Personal- ized Image Generation, 2024

    Zecheng He, Bo Sun, Felix Xu, Haoyu Ma, Ankit Ramchan- dani, Vincent Cheung, Siddharth Shah, Anmol Kalia, Ning Zhang, Peizhao Zhang, Roshan Sumbaly, Peter Vajda, and Animesh Sinha. Imagine yourself: Tuning-Free Personal- ized Image Generation, 2024. 3

    2024

  11. [19]

    Denoising dif- fusion probabilistic models

    Jonathan Ho, Ajay Jain, and Pieter Abbeel. Denoising dif- fusion probabilistic models. In Advances in Neural Informa- tion Processing Systems 33: Annual Conference on Neural Information Processing Systems 2020, NeurIPS 2020, De- cember 6-12, 2020, virtual, 2020. 2

    2020

  12. [20]

    On exact inversion of dpm-solvers

    Seongmin Hong, Kyeonghyun Lee, Suh Yoon Jeon, Hyewon Bae, and Se Young Chun. On exact inversion of dpm-solvers. CoRR, abs/2311.18387, 2023. 3

    2023 arXiv

  13. [21]

    Image quality metrics: Psnr vs

    Alain Hor ´e and Djemel Ziou. Image quality metrics: Psnr vs. ssim. In 2010 20th International Conference on Pattern Recognition, pages 2366–2369, 2010. 7

    2010

  14. [22]

    Hu, Yelong Shen, Phillip Wallis, Zeyuan Allen- Zhu, Yuanzhi Li, Shean Wang, Lu Wang, and Weizhu Chen

    Edward J. Hu, Yelong Shen, Phillip Wallis, Zeyuan Allen- Zhu, Yuanzhi Li, Shean Wang, Lu Wang, and Weizhu Chen. Lora: Low-rank adaptation of large language models. In The Tenth International Conference on Learning Represen- tations, ICLR 2022, Virtual Event, April 25-29, 2022....

    2022

  15. [23]

    Fashion and textile design with Photoshop and Illustrator : professional creative practice

    Robert Hume. Fashion and textile design with Photoshop and Illustrator : professional creative practice . Fairchild Books, an imprint of Bloomsbury Publishing, Plc, London ;,

  16. [24]

    Open- clip, 2021

    Gabriel Ilharco, Mitchell Wortsman, Ross Wightman, Cade Gordon, Nicholas Carlini, Rohan Taori, Achal Dave, Vaishaal Shankar, Hongseok Namkoong, John Miller, Han- naneh Hajishirzi, Ali Farhadi, and Ludwig Schmidt. Open- clip, 2021. If you use this software, please cite it as below. 6

    2021

  17. [25]

    EGAIN: extended gan inversion

    Wassim Kabbani, Marcel Grimmer, and Christoph Busch. EGAIN: extended gan inversion. In 10th European Work- shop on Visual Information Processing, EUVIP 2022, Lis- bon, Portugal, September 11-14, 2022 , pages 1–6. IEEE,

    2022

  18. [26]

    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 Advances in Neural Information Processing Sys- tems 35: Annual Conference on Neural Information Process- ing Systems 2022, NeurIPS 2022, New Orleans, ...

    2022

  19. [27]

    Towards understanding cross and self-attention in stable diffusion for text-guided image editing

    Bingyan Liu, Chengyu Wang, Tingfeng Cao, Kui Jia, and Jun Huang. Towards understanding cross and self-attention in stable diffusion for text-guided image editing. In IEEE/CVF Conference on Computer Vision and Pattern Recognition, CVPR 2024, Seattle, WA, USA, June 16-22, 2024 ,...

    2024

  20. [28]

    Dpm-solver++: Fast solver for guided sampling of diffusion probabilistic models

    Cheng Lu, Yuhao Zhou, Fan Bao, Jianfei Chen, Chongx- uan Li, and Jun Zhu. Dpm-solver++: Fast solver for guided sampling of diffusion probabilistic models. CoRR, abs/2211.01095, 2022. 7

    2022 arXiv

  21. [29]

    Smart Fash- ion: A review of AI applications in the fashion & apparel industry, 2021

    Seyed Omid Mohammadi and Ahmad Kalhor. Smart Fash- ion: A review of AI applications in the fashion & apparel industry, 2021. 1

    2021

  22. [30]

    Null-text inversion for editing real images using guided diffusion models

    Ron Mokady, Amir Hertz, Kfir Aberman, Yael Pritch, and Daniel Cohen-Or. Null-text inversion for editing real images using guided diffusion models. In IEEE/CVF Conference on Computer Vision and Pattern Recognition, CVPR 2023, Van- couver, BC, Canada, June 17-24, 2023 , pages 60...

    2023

  23. [31]

    Drag your GAN: interactive point-based manipulation on the generative image manifold

    Xingang Pan, Ayush Tewari, Thomas Leimk ¨uhler, Lingjie Liu, Abhimitra Meka, and Christian Theobalt. Drag your GAN: interactive point-based manipulation on the generative image manifold. In ACM SIGGRAPH 2023 Conference Pro- ceedings, SIGGRAPH 2023, Los Angeles, CA, USA, August...

    2023

  24. [32]

    Richter, Christo- pher Pal, and Marc Aubreville

    Pablo Pernias, Dominic Rampas, Mats L. Richter, Christo- pher Pal, and Marc Aubreville. W ¨urstchen: An efficient architecture for large-scale text-to-image diffusion models. In The Twelfth International Conference on Learning Rep- resentations, ICLR 2024, Vienna, Austria, May...

    2024

  25. [33]

    FICE: text-conditioned fashion image editing with guided GAN inversion

    Martin Pernus, Clinton Fookes, Vitomir Struc, and Simon Dobrisek. FICE: text-conditioned fashion image editing with guided GAN inversion. CoRR, abs/2301.02110, 2023. 3

    2023 arXiv

  26. [34]

    Ramesh, Gabriel Goh, Sandhini Agarwal, Girish Sastry, Amanda Askell, Pamela Mishkin, Jack Clark, Gretchen Krueger, and Ilya Sutskever

    Alec Radford, Jong Wook Kim, Chris Hallacy, A. Ramesh, Gabriel Goh, Sandhini Agarwal, Girish Sastry, Amanda Askell, Pamela Mishkin, Jack Clark, Gretchen Krueger, and Ilya Sutskever. Learning transferable visual models from nat- ural language supervision. In ICML, 2021. 6

    2021

  27. [35]

    Artificial intelligence and sus- tainability in the fashion industry: a review from 2010 to

    Leo Ramos, Francklin Rivas-Echeverr ´ıa, Anna Gabriela P´erez, and Edmundo Casas. Artificial intelligence and sus- tainability in the fashion industry: a review from 2010 to

    2010

  28. [36]

    Girshick, Piotr Doll ´ar, and Christoph Feichtenhofer

    Nikhila Ravi, Valentin Gabeur, Yuan-Ting Hu, Ronghang Hu, Chaitanya Ryali, Tengyu Ma, Haitham Khedr, Roman R¨adle, Chlo ´e Rolland, Laura Gustafson, Eric Mintun, Junt- ing Pan, Kalyan Vasudev Alwala, Nicolas Carion, Chao- Yuan Wu, Ross B. Girshick, Piotr Doll ´ar, and Christop...

    2024 arXiv

  29. [37]

    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. InIEEE/CVF Conference on Computer Vision and Pattern Recognition, CVPR 2022, New Orleans, LA, USA, June 18-24, 2022 , pages 10674–...

    2022

  30. [38]

    U-net: Convolutional networks for biomedical image segmentation

    Olaf Ronneberger, Philipp Fischer, and Thomas Brox. U-net: Convolutional networks for biomedical image segmentation. In Medical Image Computing and Computer-Assisted Inter- vention - MICCAI 2015 - 18th International Conference Mu- 10 nich, Germany, October 5 - 9, 2015, Proceed...

    2015

  31. [39]

    LAION-5b: An open large-scale dataset for train- ing next generation image-text models

    Christoph Schuhmann, Romain Beaumont, Richard Vencu, Cade W Gordon, Ross Wightman, Mehdi Cherti, Theo Coombes, Aarush Katta, Clayton Mullis, Mitchell Worts- man, Patrick Schramowski, Srivatsa R Kundurthy, Katherine Crowson, Ludwig Schmidt, Robert Kaczmarczyk, and Jenia Jitsev....

    2022

  32. [40]

    GAN inversion with semantic segmentation map for image editing

    Chang Jong Shin and Yong Seok Heo. GAN inversion with semantic segmentation map for image editing. In 13th In- ternational Conference on Information and Communication Technology Convergence, ICTC 2022, Jeju Island, Korea, Republic of, October 19-21, 2022 , pages 927–931. IEEE,

    2022

  33. [41]

    Very deep con- volutional networks for large-scale image recognition

    Karen Simonyan and Andrew Zisserman. Very deep con- volutional networks for large-scale image recognition. In 3rd International Conference on Learning Representations, ICLR 2015, San Diego, CA, USA, May 7-9, 2015, Confer- ence Track Proceedings, 2015. 3

    2015

  34. [42]

    Editing out-of-domain GAN inversion via differential activations

    Haorui Song, Yong Du, Tianyi Xiang, Junyu Dong, Jing Qin, and Shengfeng He. Editing out-of-domain GAN inversion via differential activations. InComputer Vision - ECCV 2022 - 17th European Conference, Tel Aviv, Israel, October 23-27, 2022, Proceedings, Part XVII, pages 1–17. Sp...

    2022

  35. [43]

    Denois- ing diffusion implicit models

    Jiaming Song, Chenlin Meng, and Stefano Ermon. Denois- ing diffusion implicit models. In 9th International Con- ference on Learning Representations, ICLR 2021, Virtual Event, Austria, May 3-7, 2021 . OpenReview.net, 2021. 2, 3

    2021

  36. [44]

    Kingma, Ab- hishek Kumar, Stefano Ermon, and Ben Poole

    Yang Song, Jascha Sohl-Dickstein, Diederik P. Kingma, Ab- hishek Kumar, Stefano Ermon, and Ben Poole. Score-based generative modeling through stochastic differential equa- tions. In 9th International Conference on Learning Rep- resentations, ICLR 2021, Virtual Event, Austria, May 3-7,

    2021

  37. [45]

    ESRGAN: enhanced super-resolution generative adversarial networks

    Xintao Wang, Ke Yu, Shixiang Wu, Jinjin Gu, Yihao Liu, Chao Dong, Yu Qiao, and Chen Change Loy. ESRGAN: enhanced super-resolution generative adversarial networks. In Computer Vision - ECCV 2018 Workshops - Munich, Ger- many, September 8-14, 2018, Proceedings, Part V, pages 63–

    2018

  38. [46]

    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 Transactions on Image Processing , 13(4): 600–612, 2004. 7

    2004

  39. [47]

    Tolerantgan: Text-guided image ma- nipulation tolerant to real-world image

    Yuto Watanabe, Ren Togo, Keisuke Maeda, Takahiro Ogawa, and Miki Haseyama. Tolerantgan: Text-guided image ma- nipulation tolerant to real-world image. IEEE Open Journal of Signal Processing, 5:150–159, 2024. 2

    2024

  40. [48]

    GAN inversion: A survey

    Weihao Xia, Yulun Zhang, Yujiu Yang, Jing-Hao Xue, Bolei Zhou, and Ming-Hsuan Yang. GAN inversion: A survey. IEEE Trans. Pattern Anal. Mach. Intell. , 45(3):3121–3138,

  41. [49]

    Florence-2: Advancing a unified representation for a variety of vision tasks

    Bin Xiao, Haiping Wu, Weijian Xu, Xiyang Dai, Houdong Hu, Yumao Lu, Michael Zeng, Ce Liu, and Lu Yuan. Florence-2: Advancing a unified representation for a variety of vision tasks. CoRR, abs/2311.06242, 2023. 3, 6, 1

    2023 arXiv

  42. [50]

    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. CoRR, abs/2308.06721, 2023. 2, 4, 6

    2023 arXiv

  43. [51]

    High-fidelity image inpainting with GAN inversion

    Yongsheng Yu, Libo Zhang, Heng Fan, and Tiejian Luo. High-fidelity image inpainting with GAN inversion. InCom- puter Vision - ECCV 2022 - 17th European Conference, Tel Aviv, Israel, October 23-27, 2022, Proceedings, Part XVI , pages 242–258. Springer, 2022. 3

    2022

  44. [52]

    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 IEEE/CVF International Conference on Computer Vision, ICCV 2023, Paris, France, October 1-6, 2023, pages 3813–

    2023

  45. [53]

    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. 7

    2018

  46. [54]

    In-domain GAN inversion for faith- ful reconstruction and editability

    Jiapeng Zhu, Yujun Shen, Yinghao Xu, Deli Zhao, Qifeng Chen, and Bolei Zhou. In-domain GAN inversion for faith- ful reconstruction and editability. IEEE Trans. Pattern Anal. Mach. Intell., 46(5):2607–2621, 2024. 3

    2024

  47. [55]

    Loopnet for fine-grained fashion attributes editing

    Xingxing Zou, Shumin Zhu, and Wai Keung Wong. Loopnet for fine-grained fashion attributes editing. Expert Systems with Applications, 259:125182, 2025. 3 11 Training-Free Consistency Pipeline for Fashion Repose Supplementary Material

    2025

  48. [60]

    Specifi- cally, we include further implementation details and more examples to validate our work

    Overview This supplementary material provides additional informa- tion on the proposed F ASHION REPOSE pipeline. Specifi- cally, we include further implementation details and more examples to validate our work. The supplementary material is structured as follows: • Additional ...

  49. [61]

    A photo of an opened sleeves clothing with 45-degree arm-torso an- gle

    Additional Implementation Details This section presents a detailed description of the prompts employed in our study, allowing for precise control over the generated output. The text prompts are utilized both in the baselines and in different stages of our pipeline. 7.1. Baseli...

  50. [62]

    Additional Qualitative Examples In this section, we present additional qualitative examples that illustrate the efficacy of the FASHION REPOSE pipeline in achieving consistent garment pose normalization. The examples demonstrate that our approach preserves garment identity, te...

  51. [63]

    The following figures highlight the impact of each stage on the quality of the final output

    Additional Qualitative Ablation Study This section provides additional qualitative examples to il- lustrate the contribution of each stage of the pipeline. The following figures highlight the impact of each stage on the quality of the final output. Figure 19 presents an ablati...

  52. [79]

    Springer, 2018. 6, 7

    2018

  53. [2021]

    OpenReview.net, 2021. 7

    2021

  54. [2022]

    SN Applied Sciences, 5(12):387, 2023. 2

    2023

  55. [3824]

    2, 4, 5, 6, 7, 1, 3

    IEEE, 2023. 2, 4, 5, 6, 7, 1, 3

    2023

Pith tools

Reviewed August 10, 2026 · model on record in the stance chip above.