Pith. sign in

REVIEW 4 major objections 4 minor 73 references

AI Tailoring: Evaluating Influence of Image Features on Fashion Product Popularity

T0 review · 4 major / 4 minor · reviewed 2026-08-12 · deepseek-v4-flash

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

Pith's one-line read A sales-and-frequency score can identify which fashion design features move product popularity, and the paper confirms it by AI-editing images and asking human raters.

desk verdict A promising pipeline undermined by unverified counterfactual edits and a largely circular validation; the influence score remains unproven. read the letter →

arxiv 2411.14737 v1 pith:DJ45KM6B submitted 2024-11-22 cs.CV cs.LG

Xiaomin Li , Junyi Sha This is my paper · ORCID
classification cs.CVcs.LG
keywords fashionimageanalysisinfluencescoredemandforecastingdiffusion-basededitingablationstudyhumanpreferencesurveysalesclassclassificationfeatureimportanceranking
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 tries to establish that the design features written in a fashion product's caption can be ranked by how much they matter to buyers, using only historical sales data and how often each feature appears. It proposes an influence score for each feature, builds a classifier called the Fashion Demand Predictor (FDP) that predicts a product's sales class from its image, text, and metadata with over 80 percent accuracy, and then uses text-prompted image editing to remove top-ranked and bottom-ranked features from real product photos. The authors argue that the ablation results—removing high-scoring features lowers predicted popularity while removing low-scoring features raises it—agree with human preference surveys, so the score is a usable guide for design and marketing decisions. If right, the framework gives fashion teams an automated, data-driven way to decide which features to keep, add, or drop before a product ships.

What carries the argument

The central object is the influence score, Eq. (2): $$\operatorname{Influence}(f_i) = \frac{1}{|S_i|}\sum_{s\in S_i} N_S(s) + \$\lambda$ \cdot N_P(p_i),$$ the average normalized sales of products containing feature $f_i$ plus a frequency-regularization term weighted by $\lambda = 0.15$. The score converts sales history into a per-feature popularity label that ranks 1,147 features from 'good' to 'bad.' The validation machinery is the Fashion Demand Predictor (FDP), a Random Forest classifier that maps image embeddings (FashionCLIP), text embeddings (Sentence-BERT), and tabular product attributes to one of three equal-quantile sales classes; its softmax probabilities are turned into a scalar popularity score via Eq. (4). The third piece is the ablation loop: diffusion-based editors remove a targeted feature from an original product image, and the change in FDP score and in human preference between original and edited versions is used to confirm the feature's influence.

What would settle it

Run a controlled ablation where the edit is verified to change only the target feature—e.g., a pixel-level or perceptual similarity test showing the rest of the image is unchanged—and check whether the influence-score ranking still predicts the direction of FDP and human preference changes; if it does not, the causal claim fails. A quicker check: permute feature-sales associations and recompute influence scores; if random 'good'/'bad' labels predict human preferences as well as the real ones, the score is capturing catalog correlation rather than causal influence.

Watch

Extended reading notes

Core claim

At the heart of the paper is the influence score for each of 1,147 caption-derived design features: $$\operatorname{Influence}(f_i) = \frac{1}{|S_i|}\sum_{s\in S_i} N_S(s) + \$\lambda$ \cdot N_P(p_i),$$ with $\lambda = 0.15$, where the first term is the average min-max-normalized sales of products containing the feature and the second penalizes rarely seen features. High-scoring features are labeled 'good' and low-scoring 'bad.' To validate the ranking, the authors train the Fashion Demand Predictor (FDP), a Random Forest classifier that takes FashionCLIP image embeddings, Sentence-BERT caption embeddings, and 36 categorical or numeric product attributes and assigns each product to one of three equal-quantile sales classes with 81% test accuracy. They then remove a targeted feature from real product images with two diffusion-based editors, InstructPix2pix-Distill and Adobe Firefly, and compare the original and edited versions using the FDP's scalar popularity score $s_j = \sum_{i=1}^{3} P(C_{ji})\cdot i$ and human preference surveys. The paper reports that removing 'good' features lowers FDP scores in all nine tested cases and human preference in seven of nine, while removing 'bad' features raises FDP scores in seven of nine and human preference in five of nine; this agreement is taken as evidence that the influence score captures features that genuinely drive demand.

Load-bearing premise

The argument depends on the image edits removing exactly the targeted feature and nothing else, because the paper provides no fidelity check that the edited images differ only in that feature; if the edits change style, color, or other attributes, the ablation results do not isolate the feature's influence.

Editorial extensions

If this is right

  • Designers could rank candidate features by influence score before production, using only the historical catalog's sales and captions, and use the ranking to prioritize high-scoring features and drop low-scoring ones.
  • The Fashion Demand Predictor offers a fast, cheap proxy for market response: training takes about 15 minutes on a CPU over 8,503 products, compared with several GPU-hours to fine-tune a multimodal LLM.
  • The same ablation pipeline can be reused as a quality check for AI-edited product images: if removing a supposedly 'good' feature does not lower the FDP score, the edit or the score needs scrutiny.
  • The method is automated end-to-end—caption cleaning, synonym clustering, scoring, and prediction—so it can be rerun as new sales data arrive, keeping the feature ranking current.

Reading between the lines

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

  • The influence score is a frequency-adjusted conditional average of past sales, so it describes which features the training catalog's customers bought, not why they bought; the causal reading rests entirely on how clean the diffusion edits are.
  • The same score-and-ablate recipe should transfer to any product domain with attribute text and sales data—furniture, consumer electronics, packaged goods—where the limiting step is editing images so that only one attribute changes.
  • Replacing the fixed $\lambda = 0.15$ with an empirical-Bayes shrinkage that downweights high-variance rare features could stabilize the ranking of features seen in only a handful of products, which the current frequency penalty only partially addresses.
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 / 4 minor

Summary. The paper proposes an 'influence score' for fashion product features, defined as the average min-max normalized sales of products containing the feature plus a frequency regularization term (Eq. 2), and a Fashion Demand Predictor (FDP) that classifies products into three sales classes from image, text, and tabular data. The authors validate the influence score through an ablation study in which a small set of 'good' and 'bad' features are removed from product images using diffusion models, and the resulting images are scored by the FDP and by human raters. The FDP reaches 81% accuracy on the 3-class sales task, and the paper claims that removing good features lowers predicted popularity while removing bad features raises it. The paper also compares FDP with a LLaVA baseline on a triplet ranking task.

Significance. The problem is timely and the idea of leveraging sales data to score visual features is appealing. The FDP model's accuracy is decent, and the combination of automated feature scoring with image editing is an interesting direction. However, the validation of the central influence score is not convincing: it relies on internal consistency with the FDP, which is trained on the same sales labels, and on a human study that shows only weak support for the 'bad' feature removal. The paper also provides no evidence that the diffusion edits isolate the targeted feature. If the influence score were validated with independent held-out data and faithful edits, it could be a useful contribution; as presented, the evidence does not support the claims.

major comments (4)
  1. [§3.3, Eq. (2); §5.1, Experiment 2] The influence score and the FDP are both computed from the same sales labels: the former as the average normalized sales of products containing a feature, the latter trained on quantile-based sales classes. Table 3 therefore tests whether two models of the same target agree; it cannot serve as an independent validation of the influence score. The human survey in Table 4 is independent, but for 'bad' features the modified image is preferred in only 5 of 9 cases (with 1 tie), which is close to chance and does not establish the claim that the influence score identifies features whose removal improves popularity.
  2. [§4; Tables 3–4] The ablation assumes that InstructPix2pix-Distill and Adobe Firefly remove exactly the targeted feature and leave all other attributes unchanged. The paper provides no fidelity check (e.g., an attribute classifier or human annotation verifying that only the target changed). The text itself notes that Firefly 'can generate content that does not fit well with the rest of the product' (Section 4), and the appendix examples (Figures 4–6) show apparent changes in style and other attributes. Without such verification, the score changes in Tables 3 and 4 cannot be attributed to the intended feature.
  3. [§5.2, Tables 3–4] The evaluation uses only 9 features per side, each with a single product. No error bars, confidence intervals, or significance tests are reported. In Table 3, some differences are very small (e.g., 'Stripped Finish': 1.345 vs 1.342), and in Table 4 the 'bad' feature results are mixed (5 AI, 3 original, 1 tie). This sample size and effect size do not support the strong claim that the influence score reliably identifies good and bad features.
  4. [§3.3, Eq. (2)] The regularization term λ·NP(pi) is added after min-max normalization of feature frequencies, so higher-frequency features receive a larger positive increment; yet the text states that a positive λ 'penalizes features with lower frequencies.' This is at best a relative penalty, and the choice λ=0.15 is not justified by any sensitivity analysis. Since the influence score is the central quantity, its definition should be clarified and its hyperparameters analyzed.
minor comments (4)
  1. [Table 3 and Figure 6] 'Stripped Finish' appears in the table but the figure caption and the surrounding text refer to a 'striped pattern'; the spelling should be consistent ('Striped Finish').
  2. [§3.1 and §3.4] The paper states there are 22 categorical and 13 numerical features in Section 3.1, but later refers to '36 categorical and numeric product data features' in Section 3.4; the discrepancy should be resolved.
  3. [Abstract and §5] The abstract says products are 'enhanced with' good features, but the experiments only remove features; this wording is misleading and should be corrected.
  4. [§5.1, Eq. (5)] The Kendall tau formula sums over triples with a constant denominator; the notation could be simplified for readability, but this does not affect the results.

Circularity Check

1 steps flagged · score 4.0 of 10

Partial circularity: FDP-based validation of the influence score shares the same sales labels; the human survey supplies independent but partial support.

  1. fitted input called prediction [Section 3.3 Eq. (2); Section 3.4.1 Eq. (3); Section 5.1 Experiment 2; Table 3]
    "Influence(fi) def = 1/|Si| Σ_{s∈Si} N_S(s) + λ · N_P(pi). ... The correct sales class label, denoted as yi, is derived by dividing the continuous true sales data into k quantile-based classes. ... we select 9 'good' features and 9 'bad' features according to their influence score defined in 2. For each feature, we select one associated product and conduct an ablation by removing that feature from the original product. Then and evaluate how the modification affects the product's popularity prediction of our FDP model."

    Eq. (2) defines the influence score as the mean normalized sales of products carrying the feature, plus a frequency term. Eq. (3) trains FDP to predict sales classes derived from exactly those sales values. Experiment 2 then uses FDP's predicted popularity shift after feature removal as evidence that the influence score identifies causally impactful features. Any model trained on the same sales labels will tend to move its prediction in the direction of the feature's average sales, so the observed agreement is a consistency check between two functions of the same target, not an independent confirmation that the feature causally drives popularity.

full rationale

The paper's central derivation chain is not self-referential by equation: the influence score (Eq. 2) is a caption-feature aggregate of normalized sales, and FDP (Eq. 3) is a separate classifier trained on quantized sales labels; neither is defined in terms of the other. However, Experiment 2 uses FDP's predicted popularity as evidence that the influence score identifies impactful features. Since both the influence score and FDP's training target are computed from the same sales data, the FDP agreement is an internal-consistency result rather than an independent confirmation; it is not a forced equality, but it is a same-label check. The human preference survey is independent and partially supports the conclusion, so the central claim does not reduce entirely to its inputs. The paper's own caveat that Adobe Firefly 'can generate content that does not fit well with the rest of the product' is a confound for the causal interpretation of the ablation, but that is a correctness or validity risk, not circularity. Self-citations ([54], [34]) are present but not load-bearing: they supply background and a methodological analogy, not the core derivation. No uniqueness theorem or ansatz is smuggled in via citation. Overall score 4 reflects one partial same-label validation step while acknowledging the independent human-survey evidence.

Assumptions & free parameters 4 free parameters · 6 assumptions · 0 invented entities

The central claim rests on several domain assumptions about the proxy relationship between sales and popularity, the fidelity of caption-to-feature mapping, and the counterfactual validity of diffusion edits. The only free parameters are the influence score regularizer λ, the clustering threshold τ0, the MinHash dimension d, and the number of sales classes C, all set by hand or by accuracy without sensitivity analysis. No new physical or conceptual entities are introduced.

free parameters (4)
  • λ (influence score regularizer) = 0.15
    Eq. (2) adds λ times the normalized feature frequency to the mean normalized sales. The paper states λ=0.15 is set to balance the two terms, with no sensitivity analysis and with tuning left to future work.
  • τ0 (MinHashLSH similarity threshold) = 0.8
    Section 3.2 sets the clustering threshold τ0=0.8 to group synonymous feature phrases; changing it would alter the final feature set.
  • d (MinHash encoding dimension) = 128
    Section 3.2 chooses d=128 for the MinHash signature dimension; this is a hyperparameter of the feature deduplication step.
  • C (number of sales classes) = 3
    Table 1 shows accuracy drops as C increases. The paper selects C=3 because it gives the highest accuracy, making the number of classes a performance-driven choice.
assumptions (6)
  • domain assumption Sales is a valid proxy for product popularity.
    Section 3.4 states this assumption and cites reference [41]; it is load-bearing because the influence score and the FDP training labels both rest on it.
  • domain assumption Caption phrases correspond to visually identifiable design features.
    The influence score is computed on caption-derived features, and Experiment 2 assumes those features can be visually edited in images. Section 3.2 and Section 5.1 rely on this.
  • domain assumption FashionCLIP and Sentence-BERT embeddings capture the visual and textual information relevant to popularity.
    Section 3.4 uses these embeddings as inputs to the FDP without validating that they contain the features that drive sales.
  • domain assumption MinHashLSH with GPT-4 representative selection preserves the true semantic feature set.
    Section 3.2 collapses 40+ synonym groups into 1,147 features; errors in clustering or representative selection propagate into the influence scores.
  • domain assumption The diffusion models edit the image by removing only the targeted feature, without introducing confounding visual changes.
    Section 4 acknowledges editing difficulties but does not verify fidelity; the ablation results in Section 5.2 depend on this assumption.
  • domain assumption The survey participants are representative of the target consumer population.
    The authors recruit female participants (100 for triplet ranking, 112 for ablation) because the products target women, but the small and self-selected Prolific sample may not generalize.

how reviews work

0 comments
Cite this review

Pith. "Pith review of AI Tailoring: Evaluating Influence of Image Features on Fashion Product Popularity." pith.science (2026). https://pith.science/paper/DJ45KM6B

@misc{pith2026241114737,
  author       = {Pith},
  title        = {Pith review of: AI Tailoring: Evaluating Influence of Image Features on Fashion Product Popularity},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/DJ45KM6B}},
  note         = {Machine review of arXiv:2411.14737}
}
read the original abstract

Identifying key product features that influence consumer preferences is essential in the fashion industry. In this study, we introduce a robust methodology to ascertain the most impactful features in fashion product images, utilizing past market sales data. First, we propose the metric called "influence score" to quantitatively assess the importance of product features. Then we develop a forecasting model, the Fashion Demand Predictor (FDP), which integrates Transformer-based models and Random Forest to predict market popularity based on product images. We employ image-editing diffusion models to modify these images and perform an ablation study, which validates the impact of the highest and lowest-scoring features on the model's popularity predictions. Additionally, we further validate these results through surveys that gather human rankings of preferences, confirming the accuracy of the FDP model's predictions and the efficacy of our method in identifying influential features. Notably, products enhanced with "good" features show marked improvements in predicted popularity over their modified counterparts. Our approach develops a fully automated and systematic framework for fashion image analysis that provides valuable guidance for downstream tasks such as fashion product design and marketing strategy development.

Figures

Figures reproduced from arXiv: 2411.14737 by the authors.

Figure 1
Figure 1. Pipeline of our methodology mathematically as: F¯ = [ k i=1 Gi . To eliminate this redundancy, we have opted to collapse each group Gi into a singleton, with only one representa￾tive element. Specifically, we select a representative feature fi from each Gi to construct our refined feature set F, ex￾pressed as: F def = {f1, f2, . . . , fk} (1) This approach ensures a more streamlined and distinct set of features for … view at source ↗
Figure 2
Figure 2. Density distribution of normalized fashion product sales. [PITH_FULL_IMAGE:figures/full_fig_p005_2.png] view at source ↗
Figure 3
Figure 3. A schematic overview o Fashion Demand Predictor(FDP): the model takes three types of input features: (1) descriptive text [PITH_FULL_IMAGE:figures/full_fig_p006_3.png] view at source ↗
Figures from the paper (3 more)
Figure 4
Figure 4. Figure 4: Comparison of Original (left) and AI-Modified (right) [PITH_FULL_IMAGE:figures/full_fig_p013_4.png]
Figure 7
Figure 7. Figure 7: Llava prompt for generating the popularity scores. [PITH_FULL_IMAGE:figures/full_fig_p013_7.png]
Figure 6
Figure 6. Figure 6: Comparison of Original (left) and AI-Modified (right) [PITH_FULL_IMAGE:figures/full_fig_p013_6.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

73 extracted references · 47 canonical work pages

  1. [1]

    Gpt-4 technical report

    Josh Achiam, Steven Adler, Sandhini Agarwal, Lama Ah- mad, Ilge Akkaya, Florencia Leoni Aleman, Diogo Almeida, Janko Altenschmidt, Sam Altman, Shyamal Anadkat, et al. Gpt-4 technical report. arXiv preprint arXiv:2303.08774 ,

    arXiv

  2. [2]

    Adobe introduces firefly image 3 foundation model to take creative exploration and ideation to new heights

    Adobe. Adobe introduces firefly image 3 foundation model to take creative exploration and ideation to new heights. https://news.adobe.com/news/news- details / 2024 / adobe - introduces - firefly - image - 3 - foundation - model - to - take - creative - exploration - and - ideation - to - new-heights, 2024. 2, 6

    work page 2024

  3. [3]

    Smart clothing recommendation system with deep learning

    Batuhan AS ¸IRO˘GLU, Mehmet Ilkay Atalay, Alkan Balkaya, Erden T ¨UZ ¨UNKAN, Mustafa Da˘gtekin, and Tolga ENSAR˙I. Smart clothing recommendation system with deep learning. In 2019 3rd International Symposium on Multidisciplinary Studies and Innovative Technologies (ISMSIT) , pages 1–4. IEEE, 2019. 2

    work page 2019

  4. [4]

    Deepseek llm: Scaling open- source language models with longtermism

    Xiao Bi, Deli Chen, Guanting Chen, Shanhuang Chen, Damai Dai, Chengqi Deng, Honghui Ding, Kai Dong, Qiushi Du, Zhe Fu, et al. Deepseek llm: Scaling open- source language models with longtermism. arXiv preprint arXiv:2401.02954, 2024. 3

    arXiv 2024

  5. [5]

    Fashion dna: merging content and sales data for recommendation and article mapping

    Christian Bracher, Sebastian Heinz, and Roland V oll- graf. Fashion dna: merging content and sales data for recommendation and article mapping. arXiv preprint arXiv:1609.02489, 2016. 2

    arXiv 2016

  6. [6]

    On the resemblance and containment of documents

    Andrei Z Broder. On the resemblance and containment of documents. In Proceedings. Compression and Complexity of SEQUENCES 1997 (Cat. No. 97TB100171) , pages 21–29. IEEE, 1997. 3

    work page 1997

  7. [7]

    In- structpix2pix: Learning to follow image editing instructions

    Tim Brooks, Aleksander Holynski, and Alexei A Efros. In- structpix2pix: Learning to follow image editing instructions. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pages 18392–18402, 2023. 6

    2023

  8. [8]

    A framework for robust feature selection for real- time fashion style recommendation

    Xiaofei Chao, Mark J Huiskes, Tommaso Gritti, and Calina Ciuhu. A framework for robust feature selection for real- time fashion style recommendation. InProceedings of the 1st international workshop on Interactive multimedia for con- sumer electronics, pages 35–42, 2009. 2

    work page 2009

Show all 73 references
  1. [9]

    Vit- v-net: Vision transformer for unsupervised volumetric med- ical image registration

    Junyu Chen, Yufan He, Eric C Frey, Ye Li, and Yong Du. Vit- v-net: Vision transformer for unsupervised volumetric med- ical image registration. arXiv preprint arXiv:2104.06468 ,

    arXiv

  2. [10]

    Fashionclip: Connecting lan- guage and images for product representations

    Patrick John Chia, Giuseppe Attanasio, Federico Bianchi, Silvia Terragni, Ana Rita Magalh˜aes, Diogo Goncalves, Ciro Greco, Jacopo Tagliabue, et al. Fashionclip: Connecting lan- guage and images for product representations. 2022. 2, 5

    2022

  3. [11]

    Fash- iongan: Display your fashion design using conditional gen- erative adversarial nets

    Yi Rui Cui, Qi Liu, Cheng Ying Gao, and Zhongbo Su. Fash- iongan: Display your fashion design using conditional gen- erative adversarial nets. InComputer Graphics Forum, pages 109–119. Wiley Online Library, 2018. 1

    2018

  4. [12]

    Research and implementation of per- sonalized clothing recommendation algorithm

    Qianqian Deng, Ruomei Wang, Zixiao Gong, Guifeng Zheng, and Zhuo Su. Research and implementation of per- sonalized clothing recommendation algorithm. In 2018 7th International Conference on Digital Home (ICDH) , pages 219–223. IEEE, 2018. 2

    2018

  5. [13]

    Diffusion models beat gans on image synthesis

    Prafulla Dhariwal and Alexander Nichol. Diffusion models beat gans on image synthesis. Advances in neural informa- tion processing systems, 34:8780–8794, 2021. 2

    2021

  6. [15]

    An image is worth 16x16 words: Transformers for image recognition at scale

    Alexey Dosovitskiy, Lucas Beyer, Alexander Kolesnikov, Dirk Weissenborn, Xiaohua Zhai, Thomas Unterthiner, Mostafa Dehghani, Matthias Minderer, Georg Heigold, Syl- vain Gelly, Jakob Uszkoreit, and Neil Houlsby. An image is worth 16x16 words: Transformers for image recognition ...

    2010 arXiv

  7. [16]

    Apparel classification using convolutional neural networks

    SG Eshwar, A V Rishikesh, NA Charan, V Umadevi, et al. Apparel classification using convolutional neural networks. In 2016 international conference on ICT in business industry & government (ICTBIG), pages 1–5. IEEE, 2016. 2

    2016

  8. [17]

    A varia- tional u-net for conditional appearance and shape generation

    Patrick Esser, Ekaterina Sutter, and Bj ¨orn Ommer. A varia- tional u-net for conditional appearance and shape generation. In Proceedings of the IEEE conference on computer vision and pattern recognition, pages 8857–8866, 2018. 2

    2018

  9. [18]

    Simi- larity search in high dimensions via hashing

    Aristides Gionis, Piotr Indyk, Rajeev Motwani, et al. Simi- larity search in high dimensions via hashing. In Vldb, pages 518–529, 1999. 3

    1999

  10. [19]

    Fashionvlp: Vision language transformer for fashion re- trieval with feedback

    Sonam Goenka, Zhaoheng Zheng, Ayush Jaiswal, Rakesh Chada, Yue Wu, Varsha Hedau, and Pradeep Natarajan. Fashionvlp: Vision language transformer for fashion re- trieval with feedback. In Proceedings of the IEEE/CVF Con- ference on Computer Vision and Pattern Recognition, pages ...

    2022

  11. [20]

    Openwebtext corpus

    Aaron Gokaslan and Vanya Cohen. Openwebtext corpus. http : / / Skylion007 . github . io / OpenWebTextCorpus, 2019. 3

    2019

  12. [21]

    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 Proceedings of the IEEE conference on computer vision and pattern recognition, pages 7543–7552, 2018. 2

    2018

  13. [22]

    Denoising dif- fusion probabilistic models

    Jonathan Ho, Ajay Jain, and Pieter Abbeel. Denoising dif- fusion probabilistic models. Advances in neural information processing systems, 33:6840–6851, 2020. 2

    2020

  14. [23]

    An empirical study of llm-as-a-judge for llm evalu- ation: Fine-tuned judge models are task-specific classifiers

    Hui Huang, Yingqi Qu, Jing Liu, Muyun Yang, and Tiejun Zhao. An empirical study of llm-as-a-judge for llm evalu- ation: Fine-tuned judge models are task-specific classifiers. arXiv preprint arXiv:2403.02839, 2024. 13

    2024 arXiv

  15. [24]

    Fashionpedia: Ontology, segmentation, and an attribute localization dataset

    Menglin Jia, Mengyun Shi, Mikhail Sirotenko, Yin Cui, Claire Cardie, Bharath Hariharan, Hartwig Adam, and Serge Belongie. Fashionpedia: Ontology, segmentation, and an attribute localization dataset. In Computer Vision–ECCV 2020: 16th European Conference, Glasgow, UK, August 23...

    2020

  16. [25]

    Survey on deep learning with class imbalance

    Justin M Johnson and Taghi M Khoshgoftaar. Survey on deep learning with class imbalance. Journal of big data , 6 (1):1–54, 2019. 5

    2019

  17. [26]

    A style-based generator architecture for generative adversarial networks

    Tero Karras, Samuli Laine, and Timo Aila. A style-based generator architecture for generative adversarial networks. In Proceedings of the IEEE/CVF conference on computer vi- sion and pattern recognition, pages 4401–4410, 2019. 1

    2019

  18. [27]

    A new measure of rank correlation

    Maurice G Kendall. A new measure of rank correlation. Biometrika, 30(1-2):81–93, 1938. 6, 7

    1938

  19. [28]

    Backpropagation applied to handwrit- ten zip code recognition.Neural computation, 1(4):541–551,

    Yann LeCun, Bernhard Boser, John S Denker, Donnie Henderson, Richard E Howard, Wayne Hubbard, and Lawrence D Jackel. Backpropagation applied to handwrit- ten zip code recognition.Neural computation, 1(4):541–551,

  20. [29]

    Deep learning

    Yann LeCun, Yoshua Bengio, and Geoffrey Hinton. Deep learning. nature, 521(7553):436–444, 2015. 2

    2015

  21. [30]

    Rlaif vs

    Harrison Lee, Samrat Phatale, Hassan Mansoor, Thomas Mesnard, Johan Ferret, Kellie Ren Lu, Colton Bishop, Ethan Hall, Victor Carbune, Abhinav Rastogi, et al. Rlaif vs. rlhf: Scaling reinforcement learning from human feedback with ai feedback. In Forty-first International Confe...

  22. [31]

    Style2vec: Representation learning for fashion items from style sets

    Hanbit Lee, Jinseok Seol, and Sang-goo Lee. Style2vec: Representation learning for fashion items from style sets. arXiv preprint arXiv:1708.04014, 2017. 2

    2017 arXiv

  23. [32]

    Retrieving real world clothing images via multi-weight deep convolutional neural networks

    Ruifan Li, Fangxiang Feng, Ibrar Ahmad, and Xiaojie Wang. Retrieving real world clothing images via multi-weight deep convolutional neural networks. Cluster Computing , 22 (Suppl 3):7123–7134, 2019. 2

    2019

  24. [33]

    Alpacaeval: An automatic evaluator of instruction-following models, 2023

    Xuechen Li, Tianyi Zhang, Yann Dubois, Rohan Taori, Ishaan Gulrajani, Carlos Guestrin, Percy Liang, and Tat- sunori B Hashimoto. Alpacaeval: An automatic evaluator of instruction-following models, 2023. 13

    2023

  25. [34]

    Rule-based data selection for large language mod- els

    Xiaomin Li, Mingye Gao, Zhiwei Zhang, Chang Yue, and Hong Hu. Rule-based data selection for large language mod- els. arXiv preprint arXiv:2410.04715, 2024. 13

    2024

  26. [35]

    Generative diffusion models on graphs: Methods and applications

    Chengyi Liu, Wenqi Fan, Yunqing Liu, Jiatong Li, Hang Li, Hui Liu, Jiliang Tang, and Qing Li. Generative diffusion models on graphs: Methods and applications. arXiv preprint arXiv:2302.02591, 2023. 1

    2023 arXiv

  27. [36]

    Visual instruction tuning

    Haotian Liu, Chunyuan Li, Qingyang Wu, and Yong Jae Lee. Visual instruction tuning. Advances in neural information processing systems, 36, 2024. 2, 7

    2024

  28. [37]

    Toward ai fashion design: An attribute-gan model for cloth- ing match

    Linlin Liu, Haijun Zhang, Yuzhu Ji, and QM Jonathan Wu. Toward ai fashion design: An attribute-gan model for cloth- ing match. Neurocomputing, 341:156–167, 2019. 2

    2019

  29. [38]

    Pose guided person image gener- ation

    Liqian Ma, Xu Jia, Qianru Sun, Bernt Schiele, Tinne Tuyte- laars, and Luc Van Gool. Pose guided person image gener- ation. Advances in neural information processing systems , 30, 2017. 2

    2017

  30. [39]

    Who, where, and what to wear? extracting fashion knowledge from social media

    Yunshan Ma, Xun Yang, Lizi Liao, Yixin Cao, and Tat-Seng Chua. Who, where, and what to wear? extracting fashion knowledge from social media. In Proceedings of the 27th ACM International Conference on Multimedia , pages 257– 265, 2019. 2

    2019

  31. [40]

    Hyperspectral image classification using cnn: Application to industrial food packaging

    Leandro D Medus, Mohamed Saban, Jose V Franc´es-V´ıllora, Manuel Bataller-Mompe ´an, and Alfredo Rosado-Mu ˜noz. Hyperspectral image classification using cnn: Application to industrial food packaging. Food Control, 125:107962, 2021. 1

    2021

  32. [41]

    On- line consumers reviews: Examining the moderating effects of product type and product popularity on the review impact on sales

    Øystein Moen, Lars Jaako Havro, and Einar Bjering. On- line consumers reviews: Examining the moderating effects of product type and product popularity on the review impact on sales. Cogent Business & Management , 4(1):1368114,

  33. [42]

    A re- view on random forest: An ensemble classifier

    Aakash Parmar, Rakesh Katariya, and Vatsal Patel. A re- view on random forest: An ensemble classifier. In Inter- national conference on intelligent data communication tech- nologies and internet of things (ICICI) 2018, pages 758–763. Springer, 2019. 5

    2018

  34. [43]

    Styleclip: Text-driven manipulation of stylegan imagery

    Or Patashnik, Zongze Wu, Eli Shechtman, Daniel Cohen-Or, and Dani Lischinski. Styleclip: Text-driven manipulation of stylegan imagery. In Proceedings of the IEEE/CVF inter- national conference on computer vision , pages 2085–2094,

  35. [44]

    Fice: Text-conditioned fashion-image editing with guided gan inversion

    Martin Pernu ˇs, Clinton Fookes, Vitomir ˇStruc, and Simon Dobriˇsek. Fice: Text-conditioned fashion-image editing with guided gan inversion. Pattern Recognition , 158:111022,

  36. [45]

    Instruct-pix2pix-distill: A distillation framework for pix2pix

    quickjkee. Instruct-pix2pix-distill: A distillation framework for pix2pix. https://github.com/quickjkee/ instruct-pix2pix-distill, 2023. 2, 6

    2023

  37. [46]

    Learning transferable visual models from natural language supervi- sion

    Alec Radford, Jong Wook Kim, Chris Hallacy, Aditya Ramesh, Gabriel Goh, Sandhini Agarwal, Girish Sastry, Amanda Askell, Pamela Mishkin, Jack Clark, et al. Learning transferable visual models from natural language supervi- sion. In International conference on machine learning, ...

    2021

  38. [47]

    Zero-shot text-to-image generation

    Aditya Ramesh, Mikhail Pavlov, Gabriel Goh, Scott Gray, Chelsea V oss, Alec Radford, Mark Chen, and Ilya Sutskever. Zero-shot text-to-image generation. In International confer- ence on machine learning, pages 8821–8831. Pmlr, 2021. 1

    2021

  39. [48]

    Hierarchical text-conditional image gener- ation with clip latents

    Aditya Ramesh, Prafulla Dhariwal, Alex Nichol, Casey Chu, and Mark Chen. Hierarchical text-conditional image gener- ation with clip latents. arXiv preprint arXiv:2204.06125, 1 (2):3, 2022. 1

    2022 arXiv

  40. [49]

    Dall-e: Creating im- ages from text

    Mr D Murahari Reddy, Mr Sk Masthan Basha, Mr M Chin- naiahgari Hari, and Mr N Penchalaiah. Dall-e: Creating im- ages from text. UGC Care Group I Journal , 8(14):71–75,

  41. [50]

    Sentence-bert: Sentence embeddings using siamese bert-networks

    N Reimers. Sentence-bert: Sentence embeddings using siamese bert-networks. arXiv preprint arXiv:1908.10084 ,

    1908 arXiv

  42. [51]

    High-resolution image synthesis with latent diffusion models

    Robin Rombach, Andreas Blattmann, Dominik Lorenz, Patrick Esser, and Bj ¨orn Ommer. High-resolution image synthesis with latent diffusion models. In Proceedings of the IEEE/CVF conference on computer vision and pattern recognition, pages 10684–10695, 2022. 2

    2022

  43. [52]

    Photorealistic text-to-image diffusion models with deep 10 language understanding

    Chitwan Saharia, William Chan, Saurabh Saxena, Lala Li, Jay Whang, Emily L Denton, Kamyar Ghasemipour, Raphael Gontijo Lopes, Burcu Karagol Ayan, Tim Salimans, et al. Photorealistic text-to-image diffusion models with deep 10 language understanding. Advances in neural informat...

    2022

  44. [53]

    Design inspiration from generative net- works

    Othman Sbai, Mohamed Elhoseiny, Antoine Bordes, Y Le- Cun, and C Couprie. Design inspiration from generative net- works. Cornell University, 2018. 2

    2018

  45. [54]

    Imaged-based similarity for demand forecasting: a novel multimodal method to exploit images’ latent information

    Junyi Sha, Yuxiang Liu, Hanwei Li, David Simchi-Levi, and Michelle Xiao Wu. Imaged-based similarity for demand forecasting: a novel multimodal method to exploit images’ latent information. Available at SSRN 4817547, 2024. 4

    2024

  46. [55]

    Fashion style in 128 floats: Joint ranking and classification using weak data for feature extraction

    Edgar Simo-Serra and Hiroshi Ishikawa. Fashion style in 128 floats: Joint ranking and classification using weak data for feature extraction. InProceedings of the IEEE conference on computer vision and pattern recognition, pages 298–307,

  47. [56]

    Deep unsupervised learning using nonequilibrium thermodynamics

    Jascha Sohl-Dickstein, Eric Weiss, Niru Maheswaranathan, and Surya Ganguli. Deep unsupervised learning using nonequilibrium thermodynamics. In International confer- ence on machine learning, pages 2256–2265. PMLR, 2015. 1

    2015

  48. [57]

    Improved techniques for training score-based generative models

    Yang Song and Stefano Ermon. Improved techniques for training score-based generative models. Advances in neural information processing systems, 33:12438–12448, 2020. 1

    2020

  49. [58]

    A meta-analysis of applications of diffusion models

    Fareena Sultan, John U Farley, and Donald R Lehmann. A meta-analysis of applications of diffusion models. Journal of marketing research, 27(1):70–77, 1990. 1

    1990

  50. [59]

    Fashion im- age retrieval with text feedback by additive attention com- positional learning

    Yuxin Tian, Shawn Newsam, and Kofi Boakye. Fashion im- age retrieval with text feedback by additive attention com- positional learning. In Proceedings of the IEEE/CVF Win- ter Conference on Applications of Computer Vision , pages 1011–1021, 2023. 2

    2023

  51. [60]

    D4: Improving llm pretraining via document de- duplication and diversification

    Kushal Tirumala, Daniel Simig, Armen Aghajanyan, and Ari Morcos. D4: Improving llm pretraining via document de- duplication and diversification. Advances in Neural Infor- mation Processing Systems, 36:53983–53995, 2023. 3

    2023

  52. [61]

    An intelligent personalized fash- ion recommendation system

    Qingqing Tu and Le Dong. An intelligent personalized fash- ion recommendation system. In 2010 International Confer- ence on Communications, Circuits and Systems (ICCCAS) , pages 479–485. IEEE, 2010. 2

    2010

  53. [62]

    Texture analysis

    Mihran Tuceryan and Anil K Jain. Texture analysis. Hand- book of pattern recognition and computer vision, pages 235– 276, 1993. 2

    1993

  54. [63]

    Attention is all you need

    A Vaswani. Attention is all you need. Advances in Neural Information Processing Systems, 2017. 2

    2017

  55. [64]

    Au- tomatic attribute discovery with neural activations

    Sirion Vittayakorn, Takayuki Umeda, Kazuhiko Murasaki, Kyoko Sudo, Takayuki Okatani, and Kota Yamaguchi. Au- tomatic attribute discovery with neural activations. In Com- puter Vision–ECCV 2016: 14th European Conference, Am- sterdam, The Netherlands, October 11–14, 2016, Procee...

    2016

  56. [65]

    Texfit: Text-driven fashion image editing with diffusion models

    Tongxin Wang and Mang Ye. Texfit: Text-driven fashion image editing with diffusion models. In Proceedings of the AAAI Conference on Artificial Intelligence , pages 10198– 10206, 2024. 1

    2024

  57. [66]

    Toward intel- ligent design: An ai-based fashion designer using genera- tive adversarial networks aided by sketch and rendering gen- erators

    Han Yan, Haijun Zhang, Linlin Liu, Dongliang Zhou, Xi- aofei Xu, Zhao Zhang, and Shuicheng Yan. Toward intel- ligent design: An ai-based fashion designer using genera- tive adversarial networks aided by sketch and rendering gen- erators. IEEE Transactions on Multimedia , 25:2323–2338,

  58. [67]

    Real-time clothing recognition in surveillance videos

    Ming Yang and Kai Yu. Real-time clothing recognition in surveillance videos. In 2011 18th IEEE international con- ference on image processing, pages 2937–2940. IEEE, 2011. 2

    2011

  59. [68]

    Text-to-image synthesis for fashion design,

    Zhengrong Yi. Text-to-image synthesis for fashion design,

  60. [69]

    Representative fashion feature extraction by leveraging weakly annotated online resources

    Heming Zhang, Siyang Li, Shanshan Cai, Haoyu Jiang, and C-C Jay Kuo. Representative fashion feature extraction by leveraging weakly annotated online resources. In 2018 25th IEEE International Conference on Image Processing (ICIP), pages 2640–2644. IEEE, 2018. 2

    2018

  61. [70]

    What makes a good image? airbnb demand analytics leveraging interpretable image features

    Shunyuan Zhang, Dokyun Lee, Param Vir Singh, and Kan- nan Srinivasan. What makes a good image? airbnb demand analytics leveraging interpretable image features. Manage- ment Science, 68(8):5644–5666, 2022. 1

    2022

  62. [71]

    M6-ufc: Unifying multi-modal controls for conditional im- age synthesis via non-autoregressive generative transform- ers

    Zhu Zhang, Jianxin Ma, Chang Zhou, Rui Men, Zhikang Li, Ming Ding, Jie Tang, Jingren Zhou, and Hongxia Yang. M6-ufc: Unifying multi-modal controls for conditional im- age synthesis via non-autoregressive generative transform- ers. arXiv preprint arXiv:2105.14211, 2021. 2

    2021 arXiv

  63. [72]

    Agieval: A human-centric benchmark for evaluating foundation models

    Wanjun Zhong, Ruixiang Cui, Yiduo Guo, Yaobo Liang, Shuai Lu, Yanlin Wang, Amin Saied, Weizhu Chen, and Nan Duan. Agieval: A human-centric benchmark for evaluating foundation models. arXiv preprint arXiv:2304.06364, 2023. 13

    2023 arXiv

  64. [73]

    Edge detection techniques-an overview

    Djemel Ziou and Salvatore Tabbone. Edge detection techniques-an overview. Pattern Recognition and Image Analysis: Advances in Mathematical Theory and Applica- tions, 8(4):537–559, 1998. 2 11

    1998

  65. [74]

    good” or “bad

    Appendix 7.1. Details of Data This section provides additional context on the dataset and survey used in this study. Table 5 lists all relevant columns (13 numerical features and 22 categorical features) from the dataset, outlining the features that contributed to the analy- s...

Pith tools

Reviewed August 12, 2026 · model on record in the stance chip above.