摘 要：随着电子商务的发展，在电商平台中用户如何挑选时尚好看的服饰单品存在困难，如何从大量的服装中搭配出适合用户需求的服装，成为时尚推荐中的热门话题。在服装构成分析中，对服饰之间的兼容性关系进行建模是一个必不可少的因素。为弥补在整个兼容性分析中不会考虑同一类时尚品的相似性问题的缺陷，提出基于类间关系和类内关系的潜在类别图嵌入（LCGE）学习方法。与现有的解决方案相比，此方法使用了服饰套装和类别信息之间的视觉结构信息。通过这种方法，能够满足用户获得符合时尚美学服饰的需求，并促进购买、刺激消费。
关键词：图卷积网络；服装兼容性分析；服装搭配；融合算法
DOI:10.19850/j.cnki.2096-4706.2023.02.016
基金项目：国家自然科学基金项目（61872394）
中图分类号：TP391 文献标识码：A 文章编号：2096-4706（2023）02-0062-07
Potential Category Map Embedded Suit Matching Model for Online Dress E-Commerce Platform
WU Yunzhi 1, SHI Xiaohong2, WANG Guan3
(1.School of Economics and Management, Guangzhou Modern Information Engineering College, Guangzhou 510670, China; 2.School of Information Engineering, Guangzhou Panyu Polytechnic, Guangzhou 511487, China; 3.School of Computer Science and Engineering, Sun Yat-sen University, Guangzhou 510006, China)
Abstract: With the development of E-commerce, users have difficulty in choosing fashionable and beautiful clothing items on the E-commerce platform. How to match a large number of clothing to suit the user's needs has become a hot topic in fashion recommendation. Modeling the compatibility relationship between garments is an essential factor in garment composition analysis. In order to make up for the defect that the similarity problem of the same type of fashion is not considered in the entire compatibility analysis, we proposed a latent class graph embedding (LCGE) learning method based on inter class relations and intra class relations. Compared with the current solutions, this method uses the visual structure information between the outfit and the categories information. This method can meet the needs of users to obtain fashionable and aesthetic clothing, and promote purchase and stimulate consumption.
Keywords: graph convolution network; clothing compatibility analysis; clothing matching; aggregation algorithm
参考文献：
[1] CUCURULL G，TASLAKIAN P，VAZQUEZ D. Context aware visual compatibility prediction [C]//2019 IEEE/CVF Conference on Computer Vision and Pattern Recognition（CVPR）.Long Beach： IEEE，2019：12617-12626.
[2] LI Z K，CUI Z Y，WU S，et al. Semi supervised compatibility learning across categories for clothing matching [C]//2019 IEEE International Conference on Multimedia and Expo（ICME）. Shanghai：IEEE，2019：484-489.
[3] LIN Y L，TRAN S，DAVIS L S. Fashion outfit complementary item retrieval [C]//2020 IEEE/CVF Conference on Computer Vision and Pattern Recognition（CVPR）.Seattle：IEEE，2020：3311–3319.
[4] LI X C，WANG X，HE X N，et al. Hierarchical fashion graph network for personalized outfit recommendation [EB/OL].[2022- 08-14].https://dl.acm.org/doi/10.1145/3397271.3401080.
[5] LIN Y S，MOOSAEI M，YANG H. Outfitnet：Fashion outfit recommendation with attention-based multiple instance learning [EB/ OL].[2022-08-12].https://dl.acm.org/doi/abs/10.1145/3366423.3380096.
[6] YANG X，MA Y S，LIAO L Z，et al. TransNFCM： Translation-based neural fashion compatibility modeling [EB/OL].[2022- 08-12].https://arxiv.org/pdf/1812.10021.pdf.
[7] VASILEVA M L，PLUMMER B A，DUSAD K，et al. Learning type-aware embeddings for fashion compatibility [J/ OL].arXiv：1803.09196 [cs.CV].[2022-08-14].https://arxiv.org/ abs/1803.09196.
[8] HAN X J，SONG X M，YIN J H，et al. Prototype-guided attribute-wise interpretable scheme for clothing matching [EB/OL]. [2022-08-14].https://dl.acm.org/doi/10.1145/3331184.3331245.
[9] HAN X T，WU Z X，JIANG Y G，et al. Learning fashion compatibility with bidirectional lstms [J/OL].arXiv：1707.05691 [cs. CV].[2022-08-12].https://arxiv.org/abs/1707.05691.
[10] YANG X，DU X Y，WANG M. Learning to match on graph for fashion compatibility modeling [EB/OL].[2022-08-16].https://ojs. aaai.org/index.php/AAAI/article/view/5362.
[11] FENG Z L，YU Z Y，YANG Y Z，et al. Interpretable partitioned embedding for customized fashion outfit composition [J/ OL].arXiv：1806.04845 [cs.CV].[2022-08-16].https://arxiv.org/ abs/1806.04845v4.
[12] SONG X M，FENG F L，LIU J H，et al. Neurostylist： Neural compatibility modeling for clothing matching [EB/OL].[2022-08- 16].https://dl.acm.org/doi/10.1145/3123266.3123314.
[13] VEIT A，KOVACS B，BELL S，et al. Learning visual clothing style with heterogeneous dyadic co-occurrences [C]//2015 IEEE International Conference on Computer Vision（ICCV）.Santiago： IEEE，2015：4642–4650.
[14] YANG X W，XIE D L，WANG X，et al. Learning tuple compatibility for conditional outfit recommendation [J/OL].arXiv： 2008.08189 [cs.CV].[2022-08-18].https://arxiv.org/abs/2008.08189.
[15] LI Y C，CAO L L，ZHU J，et al.Mining fashion outfit composition using an end-to-end deep learning approach on set data [J]. IEEE Transactions on Multimedia，2017，19（8）：1946–1955.
[16] KIPF T N，WELLING M. Semi-supervised classification with graph convolutional networks [J/OL].arXiv：1609.02907 [cs.LG]. [2022-08-18].https://arxiv.org/abs/1609.02907v3.
[17] CUI Z Y，LI Z K，WU S，et al. Dressing as a whole： Outfit compatibility learning based on node wise graph neural networks [J/OL].arXiv：1902.08009 [cs.IR].[2022-08-17].https://arxiv.org/ abs/1902.08009v1.
[18] HE K M，ZHANG X Y，REN S Q，et al. Deep residual learning for image recognition[C]//2016 IEEE Conference on Computer Vision and Pattern Recognition（CVPR）.Las Vegas：IEEE，2016：770-778.
作者简介：吴韵之（1992—），女，汉族，广东广州人，经济师，硕士，研究方向：信息管理。