From Paris to Berlin: Discovering Fashion Style Influences Around the World

CVPR, pp. 10133-10142, 2020.

Cited by: 0|Bibtex|Views71|DOI:https://doi.org/10.1109/CVPR42600.2020.01015
EI
Other Links: arxiv.org|dblp.uni-trier.de|academic.microsoft.com
Weibo:
We introduced a model to quantify influence of visual fashion trends, capturing the spatio-temporal propagation of styles around the world

Abstract:

The evolution of clothing styles and their migration across the world is intriguing, yet difficult to describe quantitatively. We propose to discover and quantify fashion influences from everyday images of people wearing clothes. We introduce an approach that detects which cities influence which other cities in terms of propagating thei...More

Code:

Data:

0
Introduction
  • The clothes people wear are a function of personal factors like comfort, taste, and occasion—and wider and more subtle influences from the world around them, like changing social norms, art, the political climate, celebrities and style icons, the weather, and the mood of the city in which they live.
  • Fashion itself is an evolving phenomenon because of these changing influences.
  • What gets worn continues to change, in ways fast, slow, and sometimes cyclical.
  • Crew influence those created six months later by Everlane, and vice versa? How long does it take for certain trends favored in New York City to migrate to Austin, if they do at all? And how did the infamous cerulean sweater worn by the protagonist in the movie The
Highlights
  • The clothes people wear are a function of personal factors like comfort, taste, and occasion—and wider and more subtle influences from the world around them, like changing social norms, art, the political climate, celebrities and style icons, the weather, and the mood of the city in which they live
  • While fashion influences exist along several axes, we focus on worldwide geography to capture spatio-temporal influences
  • Starting with images of fashion garments, 1) we learn a visual style model that captures the finegrained properties common among the garments; 2) we construct style popularity trajectories by leveraging images’ temporal and spatial meta information; 3) we model the influence relations between different locations for a given visual style; 4) we introduce a forecasting model that utilizes the learned influence relations together with a coherence loss for consistent and accurate predictions of future changes in style popularity
  • We model f (·) using a multilayer perceptron (MLP) and estimate the parameters θ by minimizing the mean squared error loss: Lforecast = (yti+j 1 − f (L, I|θ))2, (5)
  • While we focus on fashion styles in this work, we find similar results when considering individual visual attributes as the fashion concept of interest
  • We introduced a model to quantify influence of visual fashion trends, capturing the spatio-temporal propagation of styles around the world
Results
  • The authors demonstrate the model’s ability to forecast styles’ popularity changes around the globe using the discovered influence relations.
  • The authors analyze the influence patterns revealed by the model between major cities, how they influence global trends, and their dynamic influence trends through time.
  • Dataset The authors evaluate the approach on the GeoStyle dataset [28] which is based on Instagram and Flickr photos showing people from 44 major cities from around the world.
  • Styles and popularity The authors use attribute predictions from [28] to represent each photo with 46 fashion attributes.
  • While the authors focus on fashion styles in this work, the authors find similar results when considering individual visual attributes as the fashion concept of interest
Conclusion
  • The authors introduced a model to quantify influence of visual fashion trends, capturing the spatio-temporal propagation of styles around the world.
  • The authors' approach integrates both influence relations and a coherence regularizer to predict future style popularity conditioned on place.
  • Both the forecasting results and the analysis of the learned influences suggest potential applications in social science, where computer vision can unlock trends that are otherwise hard to capture
Summary
  • Introduction:

    The clothes people wear are a function of personal factors like comfort, taste, and occasion—and wider and more subtle influences from the world around them, like changing social norms, art, the political climate, celebrities and style icons, the weather, and the mood of the city in which they live.
  • Fashion itself is an evolving phenomenon because of these changing influences.
  • What gets worn continues to change, in ways fast, slow, and sometimes cyclical.
  • Crew influence those created six months later by Everlane, and vice versa? How long does it take for certain trends favored in New York City to migrate to Austin, if they do at all? And how did the infamous cerulean sweater worn by the protagonist in the movie The
  • Results:

    The authors demonstrate the model’s ability to forecast styles’ popularity changes around the globe using the discovered influence relations.
  • The authors analyze the influence patterns revealed by the model between major cities, how they influence global trends, and their dynamic influence trends through time.
  • Dataset The authors evaluate the approach on the GeoStyle dataset [28] which is based on Instagram and Flickr photos showing people from 44 major cities from around the world.
  • Styles and popularity The authors use attribute predictions from [28] to represent each photo with 46 fashion attributes.
  • While the authors focus on fashion styles in this work, the authors find similar results when considering individual visual attributes as the fashion concept of interest
  • Conclusion:

    The authors introduced a model to quantify influence of visual fashion trends, capturing the spatio-temporal propagation of styles around the world.
  • The authors' approach integrates both influence relations and a coherence regularizer to predict future style popularity conditioned on place.
  • Both the forecasting results and the analysis of the learned influences suggest potential applications in social science, where computer vision can unlock trends that are otherwise hard to capture
Tables
  • Table1: Forecast errors on the GeoStyle dataset [<a class="ref-link" id="c28" href="#r28">28</a>] for seasonal and deseasonalized fashion style trajectories
  • Table2: Correlations of the discovered influence patterns with meta information about the cities
Download tables as Excel
Related work
  • Visual fashion analysis, with its challenging vision problems and direct impact on our social and financial life, presents an attractive domain for vision research. In recent years, many aspects of fashion have been addressed in the computer vision literature, ranging from learning fashion attributes [2, 3, 6, 7, 26], landmark detection [39, 41], cross-domain fashion retrieval [19, 10, 42, 23], body shape and size based fashion suggestions [31, 14, 17], virtual tryon [38, 8], clothing recommendation [25, 30, 40, 18], inferring social cues from people’s clothes [35, 32, 24], outfit compatibility [16, 12], visual brand analysis [22, 11], and discovering fashion styles [21, 36, 1, 15]. Our work opens a new avenue for visual fashion understanding: modeling influence relations in fashion directly from images.

    Statistics of styles Analyzing styles’ popularity in the past gives a window on people’s preferences in fashion. Prior work considers how the frequency of attributes (e.g., floral, neon) changed over time [37], and how trends in (non-visual) clothing meta-data changed for the two cities Manila and Los Angeles [34]. Qualitative studies suggest how collaborative filtering recommendation models can account for past temporal changes of fashion [13] or what cities exhibit strong style similarities [20]. However, all this prior work analyzes style popularity in an “after the fact” manner, and looks only qualitatively at past changes in style trends. We propose to go beyond this historical perspective to forecast future changes in styles’ popularity along with supporting quantitative evaluation.
Funding
  • UT Austin is supported in part by NSF IIS-1514118
Reference
  • Ziad Al-Halah, Rainer Stiefelhagen, and Kristen Grauman. Fashion Forward: Forecasting Visual Style in Fashion. In ICCV, 2017.
    Google ScholarLocate open access versionFindings
  • Tamara L Berg, Alexander C Berg, and Jonathan Shih. Automatic attribute discovery and characterization from noisy web data. In ECCV, 2010.
    Google ScholarLocate open access versionFindings
  • Lukas Bossard, Matthias Dantone, Christian Leistner, Christian Wengert, Till Quack, and Luc Van Gool. Apparel classification with style. In ACCV, 2012.
    Google ScholarLocate open access versionFindings
  • George EP Box, Gwilym M Jenkins, Gregory C Reinsel, and Greta M Ljung. Time series analysis: forecasting and control. John Wiley & Sons, 2015.
    Google ScholarFindings
  • Chengyao Chen, Zhitao Wang, Wenjie Li, and Xu Sun. Modeling scientific influence for research trending topic prediction. In AAAI, 2018.
    Google ScholarLocate open access versionFindings
  • Huizhong Chen, Andrew Gallagher, and Bernd Girod. Describing clothing by semantic attributes. In ECCV, 2012.
    Google ScholarLocate open access versionFindings
  • Kuanting Chen, Kezhen Chen, Peizhong Cong, Winston H Hsu, and Jiebo Luo. Who are the Devils Wearing Prada in New York City? In ACM Multimedia, 2015.
    Google ScholarLocate open access versionFindings
  • Haoye Dong, Xiaodan Liang, Xiaohui Shen, Bochao Wang, Hanjiang Lai, Jia Zhu, Zhiting Hu, and Jian Yin. Towards multi-pose guided virtual try-on network. In ICCV, 2019.
    Google ScholarLocate open access versionFindings
  • Clive WJ Granger. Investigating causal relations by econometric models and cross-spectral methods. Econometrica: Journal of the Econometric Society, pages 424–438, 1969.
    Google ScholarLocate open access versionFindings
  • M Hadi Kiapour, Xufeng Han, Svetlana Lazebnik, Alexander C Berg, and Tamara L Berg. Where to buy it: Matching street clothing photos in online shops. In ICCV, 2015.
    Google ScholarLocate open access versionFindings
  • M Hadi Kiapour and Robinson Piramuthu. Brand > logo: Visual analysis of fashion brands. In ECCV Workshops, 2018.
    Google ScholarLocate open access versionFindings
  • Xintong Han, Zuxuan Wu, Weilin Huang, Matthew R. Scott, and Larry S. Davis. Finet: Compatible and diverse fashion image inpainting. In ICCV, 2019.
    Google ScholarLocate open access versionFindings
  • Ruining He and Julian McAuley. Ups and Downs: Modeling the Visual Evolution of Fashion Trends with One-Class Collaborative Filtering. In WWW, 2016.
    Google ScholarLocate open access versionFindings
  • Shintami Chusnul Hidayati, Cheng-Chun Hsu, Yu-Ting Chang, Kai-Lung Hua, Jianlong Fu, and Wen-Huang Cheng. What dress fits me best?: Fashion recommendation on the clothing style for personal body shape. In ACM Multimedia, 2018.
    Google ScholarLocate open access versionFindings
  • Wei-Lin Hsiao and Kristen Grauman. Learning the latent look: Unsupervised discovery of a style-coherent embedding from fashion images. In ICCV, 2017.
    Google ScholarLocate open access versionFindings
  • Wei-Lin Hsiao and Kristen Grauman. Creating capsule wardrobes from fashion images. In CVPR, 2018.
    Google ScholarLocate open access versionFindings
  • Wei-Lin Hsiao and Kristen Grauman. Dressing for diverse body shapes. In CVPR, 2020.
    Google ScholarLocate open access versionFindings
  • Wei-Lin Hsiao, Isay Katsman, Chao-Yuan Wu, Devi Parikh, and Kristen Grauman. Fashion++: Minimal edits for outfit improvement. ICCV, 2019.
    Google ScholarLocate open access versionFindings
  • Junshi Huang, Rogerio S Feris, Qiang Chen, and Shuicheng Yan. Cross-domain image retrieval with a dual attributeaware ranking network. In ICCV, 2015.
    Google ScholarLocate open access versionFindings
  • Hirokatsu Kataoka, Yutaka Satoh, Kaori Abe, Munetaka Minoguchi, and Akio Nakamura. Ten-million-order human database for world-wide fashion culture analysis. In CVPR Workshops, 2019.
    Google ScholarLocate open access versionFindings
  • M Hadi Kiapour, Kota Yamaguchi, Alexander C Berg, and Tamara L Berg. Hipster wars: Discovering elements of fashion styles. In ECCV, 2014.
    Google ScholarLocate open access versionFindings
  • Gunhee Kim and Eric Xing. Discovering pictorial brand associations from large-scale online image data. In ICCV Workshops, 2013.
    Google ScholarLocate open access versionFindings
  • Zhanghui Kuang, Yiming Gao, Guanbin Li, Ping Luo, Yimin Chen, Liang Lin, and Wayne Zhang. Fashion retrieval via graph reasoning networks on a similarity pyramid. In ICCV, 2019.
    Google ScholarLocate open access versionFindings
  • Iljung S Kwak, Ana Cristina Murillo, Peter N Belhumeur, David J Kriegman, and Serge J Belongie. From bikers to surfers: Visual recognition of urban tribes. In BMVC, 2013.
    Google ScholarFindings
  • Si Liu, Jiashi Feng, Zheng Song, Tianzhu Zhang, Hanqing Lu, Changsheng Xu, and Shuicheng Yan. Hi, magic closet, tell me what to wear! In ACM Multimedia, 2012.
    Google ScholarLocate open access versionFindings
  • Ziwei Liu, Ping Luo, Shi Qiu, Xiaogang Wang, and Xiaoou Tang. Deepfashion: Powering robust clothes recognition and retrieval with rich annotations. In CVPR, 2016.
    Google ScholarLocate open access versionFindings
  • Zheng Lu and Kristen Grauman. Story-driven summarization for egocentric video. In CVPR, 2013.
    Google ScholarFindings
  • Utkarsh Mall, Kevin Matzen, Bharath Hariharan, Noah Snavely, and Kavita Bala. Geostyle: Discovering fashion trends and events. In ICCV, 2019.
    Google ScholarLocate open access versionFindings
  • Kevin Matzen, Kavita Bala, and Noah Snavely. Streetstyle: Exploring world-wide clothing styles from millions of photos. arXiv preprint arXiv:1706.01869, 2017.
    Findings
  • Julian McAuley, Christopher Targett, Qinfeng Shi, and Anton Van Den Hengel. Image-based recommendations on styles and substitutes. In ACM SIGIR Conference on Research and Development in Information Retrieval, 2015.
    Google ScholarLocate open access versionFindings
  • Rishabh Misra, Mengting Wan, and Julian McAuley. Decomposing fit semantics for product size recommendation in metric spaces. In ACM Conference on Recommender Systems, 2018.
    Google ScholarLocate open access versionFindings
  • Ana C Murillo, Iljung S Kwak, Lubomir Bourdev, David Kriegman, and Serge Belongie. Urban tribes: Analyzing group photos from a social perspective. In CVPR Workshops, 2012.
    Google ScholarLocate open access versionFindings
  • Dafna Shahaf and Carlos Guestrin. Connecting the dots between news articles. In ACM SIGKDD International Conference on Knowledge Discovery and Data Mining, 2010.
    Google ScholarLocate open access versionFindings
  • Edgar Simo-Serra, Sanja Fidler, Francesc Moreno-Noguer, and Raquel Urtasun. Neuroaesthetics in Fashion: Modeling the Perception of Fashionability. In CVPR, 2015.
    Google ScholarLocate open access versionFindings
  • Zheng Song, Meng Wang, Xian-sheng Hua, and Shuicheng Yan. Predicting occupation via human clothing and contexts. In ICCV, 2011.
    Google ScholarLocate open access versionFindings
  • Andreas Veit, Balazs Kovacs, Sean Bell, Julian McAuley, Kavita Bala, and Serge Belongie. Learning visual clothing style with heterogeneous dyadic co-occurrences. In ICCV, 2015.
    Google ScholarLocate open access versionFindings
  • Sirion Vittayakorn, Kota Yamaguchi, Alexander C. Berg, and Tamara L. Berg. Runway to realway: Visual analysis of fashion. In WACV, 2015.
    Google ScholarLocate open access versionFindings
  • Bochao Wang, Huabin Zheng, Xiaodan Liang, Yimin Chen, Liang Lin, and Meng Yang. Toward characteristicpreserving image-based virtual try-on network. In ECCV, 2018.
    Google ScholarLocate open access versionFindings
  • Wenguan Wang, Yuanlu Xu, Jianbing Shen, and Song-Chun Zhu. Attentive fashion grammar network for fashion landmark detection and clothing category classification. In CVPR, 2018.
    Google ScholarLocate open access versionFindings
  • Cong Yu, Yang Hu, Yan Chen, and Bing Zeng. Personalized fashion design. In ICCV, 2019.
    Google ScholarLocate open access versionFindings
  • Weijiang Yu, Xiaodan Liang, Ke Gong, Chenhan Jiang, Nong Xiao, and Liang Lin. Layout-graph reasoning for fashion landmark detection. In CVPR, 2019.
    Google ScholarLocate open access versionFindings
  • Bo Zhao, Jiashi Feng, Xiao Wu, and Shuicheng Yan. Memory-augmented attribute manipulation networks for interactive fashion search. In CVPR, 2017.
    Google ScholarLocate open access versionFindings
Full Text
Your rating :
0

 

Tags
Comments