? 怎么样穿的更美观?服装搭配推荐
Journal of Computer Science and Technology
Quick Search in JCST
 Advanced Search 
      Home | PrePrint | SiteMap | Contact Us | Help
 
Indexed by   SCIE, EI ...
Bimonthly    Since 1986
Journal of Computer Science and Technology 2018, Vol. 33 Issue (3) :522-530    DOI: 10.1007/s11390-018-1836-1
Special Section of CVM 2018 << Previous Articles | Next Articles >>
怎么样穿的更美观?服装搭配推荐
Yu-Jie Liu, Member, CCF, Yong-Biao Gao, Ling-Yan Bian, Wen-Ya Wang, Zong-Min Li, Member, CCF
College of Computer and Communication Engineering, China University of Petroleum(East China) Qingdao 266580, China
How to wear beautiful? Clothing pair recommendation
Yu-Jie Liu, Member, CCF, Yong-Biao Gao, Ling-Yan Bian, Wen-Ya Wang, Zong-Min Li, Member, CCF
College of Computer and Communication Engineering, China University of Petroleum(East China) Qingdao 266580, China

摘要
参考文献
相关文章
Download: [PDF 1850KB]  
摘要 在本文中,我们提出了一个实用的服装推荐系统,在指定一件服装(上身或者下身服装)的情况下,该系统能自动的推荐出和指定服装相搭配的服装。然而,这是一个具有挑战性的问题因为服装种类多种多样。服装是人们外观最重要的组成信息。在我们日常生活中,人们在各种各样的场合下需要穿的得体并且美观以突显自己的自信、礼貌以及社会地位。但是如何穿着以及如何搭配服装并不是一件容易的事情。为了解决这个问题,我们提出了一个四路的卷积神经网络结构,其中一个双路网络采用siamese网络结构,训练样本是上下身服装的匹配或者不匹配对,另外一个双路网络用于学习输入图像的风格特征。本文所提出的框架可以把服装图像特征转换成两个隐式空间中,这两个隐式空间分别称为匹配空间和风格空间。在训练完这两个双路网络之后,我们使用特征距离融合的方法来融合从匹配和风格空间提取的特征。为了得到一个优化模型以及为了验证本文所提出的方法,我们扩充了一个大型的服装数据库称为“How to wear beautifully?”,简称H2WB。通过在H2WB数据集上做实验,说明了本文所提出的距离融合以及服装搭配推荐都有比较高的效率。
关键词服装推荐   服装搭配   风格学习     
Abstract: In this paper, we present a practical system to automatically suggest the most pairing clothing items, given the reference clothing (upper-body or low-body). However this has being a challenge, due to having varieties of clothing categories. Clothing is one of the most informative cues for human appearance. In our daily life, people need to wear properly and beautifully to show their confidence, politeness and social status in various occasion. But, it is not easy to decide to decide on what and how to wear at the same time to match with the selected clothes. To address this problem, we propose a quadruple network architecture, where one dual network adopts Siamese convolution neural network architecture. Training examples are pairs of upper-body and low-body clothing items that are either compatible or incompatible. Another dual convolution neural network is used to learn clothing style features of the input image. This framework allows learning a feature transformation from the images of clothing items to two latent spaces, which we called compatible space and style space. After training the two dual networks, we use a distance fusion method to fuse the features extracted from the compatible and style dual networks. To acquire an optimized model and verify our proposed method, we expand a large clothing dataset called "How to Wear Beautifully" (H2WB). Experiments on H2WB dataset demonstrated that our learning model are effective with feature distance fusion and clothing item pairing recommendation.
Keywordsclothing recommendation   clothing pair   style learning     
Received 2017-12-20;
本文基金:

This work was supported by the National Natural Science Foundation of China under Grant Nos. 61379106, 61379082 and 61227802, the Natural Science Foundation of Shandong Province of China under Grant Nos. ZR2013FM036 and ZR2015FM011, the Major Program of National Natural Science Foundation of China under Grant No. 41631073, and the Qingdao National Laboratory for Marine Science and Technology of China under Grant No. QNLM2016ORP0401.

About author: Wen-Guang Chen is a professor in the Department of Computer Science and Technology, Tsinghua University, Beijing, where he has been teaching since 2003. He received his B.S. and Ph.D. degrees both in computer science from Tsinghua University in 1995 and 2000 respectively. His research interest is in parallel and distributed computing. He is a CCF distinguished member and a CCF distinguished speaker, and an ACM member and the vice chair of ACM China Council. He has served in program committees of a variety of major conferences in the parallel and distributed computing area, including PLDI, PPoPP, SC, CGO, CCGrid, IPDPS, APSys, and ICPP. He is the editor-in-chief of Communication of ACM China Edition.
引用本文:   
Yu-Jie Liu, Yong-Biao Gao, Ling-Yan Bian, Wen-Ya Wang, Zong-Min Li.怎么样穿的更美观?服装搭配推荐[J]  Journal of Computer Science and Technology , 2018,V33(3): 522-530
Yu-Jie Liu, Yong-Biao Gao, Ling-Yan Bian, Wen-Ya Wang, Zong-Min Li.How to wear beautiful? Clothing pair recommendation[J]  Journal of Computer Science and Technology, 2018,V33(3): 522-530
链接本文:  
http://jcst.ict.ac.cn:8080/jcst/CN/10.1007/s11390-018-1836-1
Copyright 2010 by Journal of Computer Science and Technology