acceptodds
Under review as a conference paper at ICLR 2027

LayeredFashion: Recovering Complete Garment Layers through Layerwise Image Decomposition

Abstract

A clothed person is a stack of garments, yet human images are usually processed as flat pixels. We present LayeredFashion, which decomposes a clothed-person image and outfit description into an editable PSD-style stack of a variable number of named RGBA garments over a person underlay. We reverse dressing: Extraction recovers each garment, and Restoration reconstructs the clothing or body beneath it for the next step. To correct missing regions, false positives, and boundary errors, we introduce the RGB-Semantic Alpha Refiner (RSAR), combining current-image RGB, garment semantics, and whole-image context to refine alpha while preserving RGB. We train extraction and restoration LoRAs on reference and model-generated intermediate states to accommodate inherited errors. For stepwise supervision, we pair over 22K outfits with nearly 100K complete garment layers, together with person underlays and composition-consistent intermediate images. We achieve the best mean among compared general-purpose decomposition methods on all twelve garment, underlay, and recomposition metrics. The editable layers also support text-to-PSD composition and provide a basis for downstream layered 3D reconstruction.

Primary area:
applications to computer vision, audio, language, and other modalities
Keywords:
layered image decomposition, garment decomposition, human image generation, occlusion reasoning, alpha matting, generative models

Reviews

Among:
Shape:
1234this paper, Top 49%
Figure 1. Mean review rating of the reviewed papers at ICLR 2027. This paper averages 2.33 over 3 reviews, above 51% of the others.
Reviewer_JiTbOpenReview ↗
Rating
2
Confidence
3
Informativeness
2
Presentation
2
Soundness
2
Read review
Paper summary

LayeredFashion decomposes a clothed-person image and outfit description into editable garment RGBA layers and a person underlay. It reverses the specified dressing order through alternating garment extraction and image restoration. An RGB-Semantic Alpha Refiner (RSAR) corrects garment opacity, while a separate cumulative repair region controls recurrent image updates. The authors construct 22,431 outfits with 97,775 garment layers and evaluate decomposition, text-to-PSD composition, and downstream 3D reuse. The method achieves the best reported means across twelve decomposition metrics.

Critical strengths
  1. Independently editable garment layers support recomposition and asset reuse beyond flat virtual try-on or visible-region human parsing.
  2. Separating saved garment opacity from the cumulative restoration region allows earlier residuals to be repaired without enlarging the current garment layer.
  3. Seven ablations examine alpha decoding, RGB and semantic evidence, full-image correction, and recurrent updates; removing RSAR substantially increases alpha error and trajectory drift.
Critical weaknesses
  1. Performance on real clothing photographs is not sufficiently demonstrated. Quantitative evaluation mainly uses independently generated garments that are segmented and composited with source-over blending. This construction does not reproduce cross-layer shadows, physical fabric interactions, or complex real-world occlusion. The reference person underlay may itself retain clothing. Strong synthetic results therefore do not establish complete garment recovery from real photographs.

  2. Generalization to complex clothing combinations remains uncertain. The dataset uses a closed prompt pool, explicitly specified categories and order, category-based coordinate priors, and specific text-processing rules. Although the representation distinguishes inner tops from outerwear, it does not demonstrate reliable separation of multiple overlapping garments within the same category, which may instead be segmented as a single layer. This limitation may reflect both restricted training coverage and the lack of explicit completion of individual extracted garments. Tests with more complex upper-body layering, free-form descriptions, and unfamiliar garment combinations are needed.

  3. The claim of complete garment recovery is insufficiently supported. Each extracted garment combines the current image’s RGB with a refined alpha matte, without an explicit appearance-completion step for that garment. The upper-body garment layers illustrated in Figures 2 and 4 appear incomplete, particularly around occluded or rear portions.

  4. The dataset could be a valuable resource for the research community, but the paper does not explicitly commit to releasing it. Clarifying the release plan would strengthen the dataset contribution and support reproducibility.

Questions for authors

Please refer to Weakness 2 3 and 4. If the dataset will be open-source upon the acceptance of this paper. I'd like to considering raising my socre.

5 comments
AuthorsOpenReview ↗

Response to Reviewer JiTb (Part 1 / 2)

Thank you very much for your thoughtful and perceptive feedback, which has genuinely helped us further strengthen the manuscript.

W1. Real photographs: do synthetic composites capture shadows, fabric interactions, and complex occlusion; why does the underlay retain clothing?

Thank you for raising this important question. Your insights have deepened our understanding of our own work and have been particularly valuable in strengthening this work. While conducting the additional experiments you suggested, we were also pleased to discover some capabilities and generalization of our method that we had not anticipated.

(1) Real-world evaluation. Following your suggestion, we evaluated our method on real-world photographs and challenging examples involving fabric interactions (e.g., complex wrinkling in layered garments) and extreme inter-layer occlusion in Section 6.3, Figures 6 and 7. The results show that our method can not only accurately recover the layers for real-world images, but also reconstructs physically plausible wrinkles in occluded regions, with patterns consistent with pose-induced stretching. These results provide evidence beyond source-over synthetic composites.

(2) Underlay. The underlay is intentionally retained on the base body to protect bodily privacy and avoid exposing intimate regions. By default, it is excluded from target garments and not extracted as a separate layer. Its presence reflects the intended decomposition scope, rather than incomplete segmentation. We clarified this in Section 3.1 and Appendix A.1.

(3) Additional poses. We further tested a series of real captured images, more varied poses, multiple layers of similar clothing, and their combinations. These experimental results have been updated in Section 6.3, Figures 6 and 7. They demonstrate the robustness and generalization of our method to real and complex inputs.

W2. Generalization: Generalization to complex clothing combinations remains uncertain; tests with multiple overlapping garments within the same category, and free-form descriptions are needed.

Thank you for raising this concern. We have clarified the role of these design choices and added real-world examples with more complex layering and varied descriptions in Figures 6 and 7 in Section 6.3.

(1) Training structure and inference flexibility. The closed prompt pool and explicit category/order annotations are used to construct synthetic training data and provide consistent supervision. At inference, the parser derives garment categories and layer order from the supplied description, and the text-processing rules standardize this information for conditioning. Users can therefore vary the wording and the order in which garments are mentioned, as illustrated by the examples below.

The category-based coordinate priors provide coarse spatial guidance for garment localization. Localization and alpha prediction also depend on the input image, so these priors guide extraction without prescribing fixed garment masks. Inference thus retains structured conditioning while allowing descriptions beyond the training templates and using image evidence to determine the extracted layers. We have clarified this distinction in Section 4.1. Appendix E describes the parser, while Section 4.2 and Appendix C.1 detail inference conditioning.

(2) Same-category overlap. We added real-world examples in Section 6.3 with multiple overlapping T-shirts and multiple overlapping suit jackets. In these examples, the garments are successfully recovered as separate editable layers, demonstrating same-category separation in the tested cases. The scope of garment completion is discussed in our response to W3.

(3) Free-form descriptions. We invite the reviewer to examine the text prompts in the following table for results in Figures 6 and 7. These prompts vary in wording, garment-mention order, and natural sentence structure. Specifically, we can introduce clothing in different ways and can mention lower-body garments or accessories before upper-body layers. The corresponding outputs retain separate editable layers, providing qualitative evidence that the method can process these forms without requiring a fixed sentence template or garment-mention order.

FigurePromptText prompt
Figure 61A tight base-body wearing grey sweatpants, a black tee and a printed white T-shirt.
Figure 62Built up in three shirts: black long sleeves, then a white collared shirt, then a light blue short-sleeve shirt worn open--paired with dark grey trousers.
Figure 71A cream short-sleeve top tucked into a purple pleated mini skirt.
Figure 72Pink trousers and black shoes anchor the look, topped by black long sleeves, a yellow suit jacket and a red overcoat.
Figure 73All red on top of the base body: red trousers and a red top.
Figure 74A hat on top, a shirt under a suit jacket, trousers and shoes
AuthorsOpenReview ↗

Response to Reviewer JiTb (Part 2 / 2)

W3. The claim of complete garment recovery is insufficiently supported. Why do Figures 2 and 4 (now Figures 3 and 5) look incomplete?

We thank the reviewer for highlighting this ambiguity in our original wording. By “complete garment recovery,” we mean recovering the garment regions that should be visible from the input viewpoint after overlying garments are removed, while preserving the occlusion relationships required for PSD-style recomposition. This does not imply recovery of the garment’s full 360° appearance or physical geometry. We have revised the manuscript to define “complete” precisely.

(1) What is completed, and how. Our completion target is the garment appearance that should be visible from the input viewpoint after overlying garments have been removed. This includes regions occluded by outer garments in the original image, but excludes regions that remain hidden behind the body, such as rear panels behind the torso or collar surfaces hidden by the neck.

Restoration explicitly completes the garment regions exposed by removing overlying garments, before those garments are extracted. In Figure 1, for example, it reconstructs the white shirt’s previously hidden front and sleeves, as well as the upper portions of the trousers covered by the jacket. Subsequent Extraction uses this restored RGB together with the refined alpha matte. The recovered layers therefore include appearance that was occluded in the original input.

(2) Why some regions remain transparent? This is an intentional design choice consistent with our paradigm. We follow PSD-style compositing and the physical occlusion order in the input view. For the open jackets in Figures 3 and 5, rear panels behind the torso are excluded from the RGBA layer, as are collar surfaces hidden by the neck. Filling these areas would cause the garment layer to obscure the body or inner clothing that should remain visible when recomposed. To further clarify the distinction between PSD-style layer extraction and full-garment extraction, we have added Figure 2 (Section 1, p. 2), which presents each isolated coat layer alongside its recomposition over the same underlying layers. In the left pair, the PSD-style layer leaves the coat opening transparent, so the dark inner garment remains visible. In the right pair, directly overlaying the full-garment layer covers the inner garment with the completed rear panel. Our core target is therefore a recomposable, view-aligned clothing layer, rather than a reconstruction of the garment’s full physical geometry.

The recovered layers can support further completion toward physically complete garment representations, as illustrated by our downstream 3D reconstruction results in Figure 1 and Figure 9. However, this serves a different objective from PSD-style decomposition, which requires preserving the visibility and occlusion relationships of the input view.

In response to your concern, we have clarified the scope of “complete” and the intended PSD representation in Sections 1 and 3.1 and Appendix A.1, with corresponding revisions to the text and supporting figures.

W4. Dataset release: will the dataset be made available upon acceptance?

We commit to publicly releasing the complete LayeredFashion dataset, including all layer annotations and text descriptions, upon acceptance to support reproducibility and further research. This commitment is now explicitly stated in the Reproducibility statement (p. 11).

We sincerely hope that our response has addressed your concerns. Your thoughtful and perceptive feedback is sincerely appreciated and has genuinely helped us further strengthen the manuscript. Should you have any further questions, please let us know. We would be more than happy to address them and make further improvements.

Thank you for the revisions. The paper has improved since the initial submission, and the commitment to publicly release the dataset is a significant contribution. After reading the new material, I still have several concerns.

First, missing regions of outer garments and fabric occluded behind the body remain a substantial issue, as illustrated by the scarf in Figure 1, the jackets in Figures 4 and 7, and the hat in Figure 7. The claimed “complete garment recovery” appears limited to partial recovery of inner garments.

Second, all examples shown are strictly front-facing and feature no garment overlap or occlusion. Given how the method works, occlusion by the arms, such as when they are crossed in front of the body, or a side-facing pose could substantially affect decomposition quality.

Third, the real-world evaluation consists of only a few qualitative examples, which are insufficient to establish generalization in practice. Training and testing on a larger set of real photographs would strengthen the work.

For these reasons, I maintain my original Weak Reject rating.

Reviewer_15sJOpenReview ↗
Rating
3
Confidence
3
Informativeness
2
Presentation
3
Soundness
3
Read review
Paper summary

This paper presents LayeredFashion, which is a two-stage pipeline that uses a person’s image and outfit description to extract the clothing layers and then dress the person. It extracts the outermost garment layer by layer and refines alpha masks to reduce accumulated errors. The authors introduce a synthetic layered-clothing dataset and report improvements over general image-decomposition methods.

Critical strengths
  1. The authors identify the error accumulation problem in the sequential generation. They address it effectively by mixing reference images with model-generated images.

  2. The comparison with the adapted baseline shows clear improvements.

Critical weaknesses
  1. The obejctive functions are mentioned in the appendix but not in the main paper. They are an important component to understand the level of supervision required by the proposed method.

  2. The image quality of the extracted garment and the recomposed person are only evaluated using objective metrics. I suggest the authors conduct user studies on different image quality aspects to better align with human perception.

Questions for authors
  1. The paper trains the two stages sequentially, while extraction and restoration alternate during inference. Why was separate training chosen? Did the authors try joint training, and what was the main difficulty?

  2. Where do the studio model photographs come from? What selection rules are used for viewing angle, pose, existing clothing, and body occlusion?

  3. Will the collected dataset be released?

2 comments
AuthorsOpenReview ↗

Response to Reviewer 15sJ (Part 1 / 2)

We sincerely thank you for your valuable and perceptive feedback, which has truly helped us further enhance the manuscript.

W1. The objective functions should be included in the main paper to clarify the method’s supervision requirements.

We appreciate this important request. To address this issue, we have added the detailed core loss functions to Section 4.5, with full auxiliary definitions in Appendix B.2.

(1) Extraction. We train the Extraction LoRA with full-latent squared prediction error:

LE=1N∑i=1N(vθ,i−vi⋆)2.\mathcal L_E=\frac1N\sum_{i=1}^N(v_{θ,i}-v_i^⋆)^2.

Here vθ,iv_{θ,i} is the predicted diffusion velocity and vi⋆v_i^⋆ is the training velocity target derived from the reference garment latent. The reference garment is the stored layer before outfit compositing obscures it.

(2) Restoration. Let eie_i be elementwise squared prediction error and mi∈[0,1]m_i\in[0,1] the continuous current-to-target change map, broadcast over latent channels:

LR=λ1(R)1N∑iei+λ2(R)⟨e⟩m+λ3(R)⟨e⟩1−m.\mathcal L_R=\lambda_1^{(R)}\frac1N\sum_i e_i+\lambda_2^{(R)}⟨e⟩_m+\lambda_3^{(R)}⟨e⟩_{1-m}.

⟨e⟩w=∑iwieimax⁡1,∑iwi.⟨e⟩_w=\frac{\sum_i w_i e_i}{\max\\{1,\sum_i w_i\\}}.

The sums cover latent elements. The global term supervises the full prediction; the two weighted terms supervise changing and preserved regions. The reference is the lower-layer composite after garment removal. This objective precedes the scheduler’s timestep weighting; mm is a training weight, not the inference repair mask.

(3) Alpha decoder. We supervise two frozen latent sources and empty targets:

Lα=γ1(α)Lassetlatent+γ2(α)Lassetreconstruction+γ3(α)Lblank.\mathcal L_α=\gamma_1^{(α)}\mathcal L_{\rm asset}^{\rm latent}+\gamma_2^{(α)}\mathcal L_{\rm asset}^{\rm reconstruction}+\gamma_3^{(α)}\mathcal L_{\rm blank}.

The first asset term decodes cached extractor latents; the second decodes frozen-VAE latents of reference RGBA garments. Both use reference alpha/RGBA. The blank term penalizes opacity for empty targets. Asset terms supervise alpha, boundaries, and composited appearance; Appendix B.2 gives their coefficients.

(4) RSAR. For a predicted matte aa and target alpha yy, define:

Lshared(a,y)=∑r∈RωrLr(a,y)+ωBCELBCE(a,y)+ωDiceLDice(a,y).\mathcal L_{\rm shared}(a,y)=\sum_{r\in\mathcal R}ω_r\mathcal L_r(a,y)+ω_{\rm BCE}\mathcal L_{\rm BCE}(a,y)+ω_{\rm Dice}\mathcal L_{\rm Dice}(a,y).

LRSAR=Lshared(αt,y)+ηgradLgrad(αt,y)+ηdecLshared(α‾t,y).\mathcal L_{\rm RSAR}=\mathcal L_{\rm shared}(α_t,y)+η_{\rm grad}\mathcal L_{\rm grad}(α_t,y)+η_{\rm dec}\mathcal L_{\rm shared}(\overline{α}_t,y).

Each Lr\mathcal L_r is region-normalized absolute alpha error. We use ω=(4,4,4,1)ω=(4,4,4,1) for preserved foreground, missing foreground, false positives, and background. Region-balanced binary cross-entropy (BCE) supervises opacity; soft Dice supervises overlap; the gradient term supervises boundaries. Final alpha αtα_t receives all terms; intermediate alpha αˉt\barα_t receives the same loss at weight ηdec=0.5η_{\rm dec}=0.5, without the gradient term. Reference latents, composites, and alpha/RGBA targets are used only during training.

W2. The authors should conduct user studies on different image quality aspects to better align with human perception.

Thank you for recommending this perceptual evaluation, which is very helpful in improving the comprehensiveness of the evaluation. Following your suggestion, we have completed the user study and added the results to the manuscript. See Section 6.4 (p. 9) and Appendix I (Table 10, p. 36).

(1) User study protocol. We designed a questionnaire comparing our method with Qwen-Image-Layered, OmniPSD, RevealLayer, and GPT on six selected cases. All respondents completed the same 12 questions: six full-layer comparisons and six recomposition comparisons, each with two criteria. Participants chose the best result for each criterion: separation and shape, and texture and edges for layers; similarity to the original and naturalness for recomposition. Methods were labeled A–E, with assignments varying across questions. We received 33 complete responses. This yielded 792 judgments, or 198 per criterion.

(2) Results. Participants selected our method as the best in terms of separation and shape (174/198 judgments, 87.9%), texture and edges (165/198, 83.3%), recomposition similarity (168/198, 84.8%), and naturalness (171/198, 86.4%). Percentages are based on all 198 judgments for each aspect, including “No clear difference” and “Cannot judge” responses.

Table 10. Human preferences on six selected cases.

Garment layers (cols 2–3) · Recomposition (cols 4–5)

Method / responseSeparationTextureSimilarityNaturalness
Qwen-Image-Layered2 (1.0%)2 (1.0%)18 (9.1%)15 (7.6%)
OmniPSD6 (3.0%)8 (4.0%)1 (0.5%)0 (0.0%)
RevealLayer2 (1.0%)5 (2.5%)0 (0.0%)1 (0.5%)
GPT12 (6.1%)12 (6.1%)7 (3.5%)8 (4.0%)
LayeredFashion174 (87.9%)165 (83.3%)168 (84.8%)171 (86.4%)
No clear difference2 (1.0%)6 (3.0%)3 (1.5%)1 (0.5%)
Cannot judge0 (0.0%)0 (0.0%)1 (0.5%)2 (1.0%)
AuthorsOpenReview ↗

Response to Reviewer 15sJ (Part 2 / 2)

Q1. Separate training: why train sequentially despite alternating inference; was joint training tried, and what is the main difficulty?

Thank you for this important question. We tried jointly training the Extraction and Restoration LoRAs in early experiments, but the results were poor and training was unstable. Restoring the remaining garments and recovering the final person underlay both depend on accurate extraction masks. Restoration relies on the mask supplied by Extraction. If an insufficiently trained extractor includes part of the trousers in a shirt layer, the removal region has the wrong garment assignment. The resulting restoration error can propagate into later steps and contaminate the training states. Updating both LoRAs together can reinforce these errors.

Therefore, we first optimize Extraction and its alpha components, then freeze the extraction cascade while training Restoration on the imperfect states it actually produces. Fixing the upstream cascade reduces the coupling between mask prediction and image restoration, allowing Restoration to learn the extractor’s actual error patterns without simultaneously changing their source. In our experiments, this substantially improved training stability compared with joint training. The two modules still alternate at inference, with each restored image serving as the next Extraction input. This execution-time dependency is preserved without requiring the two stages to be optimized jointly. Section 4.5 and Appendix B.3 describe the training schedule. We have also clarified the benefits of sequential training in Section 4.5.

Q2. Person photographs: where do they come from, and how are viewing angle, pose, existing clothing, and body occlusion considered?

We are grateful for the opportunity to clarify the source and curation of these images.

Capturing real human subjects wearing multiple layers of clothing layer by layer in a studio would be extremely difficult, particularly in terms of maintaining pose consistency, ensuring diversity, and protecting privacy. We therefore carefully designed a data synthesis scheme.

Our dataset construction pipeline integrates image generation with human refinement and quality control at multiple stages. We use AI image-generation models to create person underlays and generate each garment layer independently on the same underlay before composing the outfit. Human curators adjust prompts and generation settings, screen natural appearance and body–garment consistency, remove backgrounds, perform segmentation, manually correct masks and contours, and reject failed samples. We cross-check candidates between the generation and verification stages. This requires repeated agent–human intervention rather than a single bulk generation pass.

The dataset primarily uses front-facing, full-body views in A-pose, and our AI-assisted garment library covers diverse categories, styles, materials, patterns, and fits. We follow the physical occlusion order of the input view in the PSD stack, leaving garment surfaces behind the body transparent. Following your comments, we have clarified the representation in Section 3.1 and Appendix A.1. Section 3.2 describes the curation workflow. Additional results added in Section 6.3 show that the model recovers garment layers in real photographs (Figure 6) and across varied poses (Figure 7).

Q3. Will the dataset be released?

Thank you for emphasizing accessibility. We commit to releasing the LayeredFashion dataset and layer annotations upon acceptance, subject to source permissions and privacy safeguards. This commitment appears in the Reproducibility statement (p. 11).

We sincerely hope that our response has addressed your concerns. We are truly grateful for your valuable and insightful comments, which have genuinely helped us further enhance our manuscript. Should you have any further questions, please let us know. We would be more than happy to address them and make further improvements.

Reviewer_CbU6OpenReview ↗
Rating
2
Confidence
3
Informativeness
2
Presentation
2
Soundness
2
Read review
Paper summary

TCFlow addresses temporal inconsistency in single-reference video color transfer by modeling the entire video clip as a unified color distribution and predicting a clip-shared rectified flow for all frames. A color aggregator is further introduced to infer the shared flow efficiently from global temporal color statistics, enabling stable and high-fidelity color transfer without optical flow or test-time optimization.

Critical strengths
  1. The paper addresses a practically important problem.
  2. The clip-shared formulation is simple and conceptually clean.
Critical weaknesses
  1. The methodological novelty of this work appears somewhat limited. A substantial part of the technical framework is closely related to prior flow-based color transfer methods, particularly ModFlows. In addition, improving temporal consistency by applying a shared color transformation across multiple video frames has already been explored in prior video color/style transfer literature.
  2. The current experiments are insufficient to demonstrate the necessity of the proposed clip-level distribution modeling. The paper argues that constructing a video-level distribution and predicting a clip-shared flow is superior to frame-wise mappings; however, the main comparison essentially contrasts frame-wise flows with the proposed clip-shared flow.
Questions for authors
  1. How would TCFlow compare with a simple baseline that estimates one ModFlows transformation from a representative frame and applies it to every frame in the clip?
  2. Can the authors clarify how much improvement originates from clip-level distribution pooling itself versus the proposed temporal Color Aggregator?
2 comments
AuthorsOpenReview ↗

Review Mismatch for Paper 21046 (LayeredFashion)

Dear Reviewer CbU6,

Thank you for contributing your time and expertise to the review process.

We are writing to bring a possible mix-up to your attention regarding the review attached to our submission, Paper 21046 (LayeredFashion). The review states, “TCFlow addresses temporal inconsistency in single-reference video color transfer.” Our submission introduces LayeredFashion, which addresses the decomposition of clothed human images into layers. Given these differences in both the method name and the research task, the review of our submission appears to have been written about another different paper.

We appreciate the substantial time and effort that reviewing requires and understand that an unintended mix-up can occur. With the rebuttal period now drawing to a close, we hope you can also understand our concern about having sufficient time to receive and respond meaningfully to feedback on our work.

Would you kindly help clarify this mismatch with the Area Chair and discuss how it might be resolved? Your help would be greatly appreciated in ensuring that our paper receives a fair evaluation and that we have an opportunity to address feedback concerning our submission.

Thank you very much for your understanding, time, and assistance. We would be happy to provide any further information that may help clarify the issue.

Best regards,

The Authors of Paper 21046 (LayeredFashion)

AuthorsOpenReview ↗

Review Mismatch for Paper 21046 (LayeredFashion)

Dear Reviewer CbU6,

With only a few hours remaining before the rebuttal deadline, we are following up on our earlier messages regarding Paper 21046 (LayeredFashion). The review still concerns a different paper, and we have not yet received a reply to our clarification. We appreciate the substantial demands on your time and understand that an inadvertent mix-up may have occurred. We would be grateful for your help in clarifying this issue with the Area Chair and finding a resolution. Thank you sincerely for your understanding, time, and assistance.

Best regards,

The Authors of Paper 21046 (LayeredFashion)

Comments on the paper

1 comment
AuthorsOpenReview ↗

Clarification of a Review Mismatch for Paper 21046 (LayeredFashion)

Dear Reviewers,

Thank you very much for your time, expertise, and careful consideration of our submission.

We would like to offer a brief clarification: the review from Reviewer CbU6 appears to concern another different paper. It discusses TCFlow and temporal inconsistency in single-reference video color transfer, whereas our paper, LayeredFashion, addresses the decomposition of clothed human images into layers.

We appreciate the considerable effort involved in reviewing and understand that an inadvertent mix-up may have occurred. We would be sincerely grateful if you could keep this mismatch in mind during the discussion, so that comments and scores from this mismatched review, which concern a different paper and are therefore not applicable to LayeredFashion, do not inadvertently influence the assessment of our submission.

Thank you again for your understanding and for the care you bring to evaluating our work. We greatly appreciate your consideration.

Best regards,

The Authors of Paper 21046 (LayeredFashion)

From OpenReview, CC BY 4.0. Authors as signed there; shown unedited.

Related papers

Loading the map…

Discussion (0)

Sign in to comment.