{"about":{"site":"https://codewithpapers.app","non_affiliation":"Code with Papers and Syntology are not affiliated with, endorsed by, or sponsored by Papers with Code, Meta, or the pwc-archive mirror.","licence":"CC BY-SA 4.0","licence_url":"https://creativecommons.org/licenses/by-sa/4.0/legalcode","attribution":"https://codewithpapers.app/attribution","modified":"archive material modified by Syntology; see the attribution page"},"url":"/paper/dressing-as-a-whole-outfit-compatibility","title":"Dressing as a Whole: Outfit Compatibility Learning Based on Node-wise Graph Neural Networks","arxiv_id":"1902.08009","date":"2019-02-21","proceeding":null,"authors":["Zeyu Cui","Zekun Li","Shu Wu","Xiao-Yu Zhang","Liang Wang"],"abstract":"With the rapid development of fashion market, the customers' demands of\ncustomers for fashion recommendation are rising. In this paper, we aim to\ninvestigate a practical problem of fashion recommendation by answering the\nquestion \"which item should we select to match with the given fashion items and\nform a compatible outfit\". The key to this problem is to estimate the outfit\ncompatibility. Previous works which focus on the compatibility of two items or\nrepresent an outfit as a sequence fail to make full use of the complex\nrelations among items in an outfit. To remedy this, we propose to represent an\noutfit as a graph. In particular, we construct a Fashion Graph, where each node\nrepresents a category and each edge represents interaction between two\ncategories. Accordingly, each outfit can be represented as a subgraph by\nputting items into their corresponding category nodes. To infer the outfit\ncompatibility from such a graph, we propose Node-wise Graph Neural Networks\n(NGNN) which can better model node interactions and learn better node\nrepresentations. In NGNN, the node interaction on each edge is different, which\nis determined by parameters correlated to the two connected nodes. An attention\nmechanism is utilized to calculate the outfit compatibility score with learned\nnode representations. NGNN can not only be used to model outfit compatibility\nfrom visual or textual modality but also from multiple modalities. We conduct\nexperiments on two tasks: (1) Fill-in-the-blank: suggesting an item that\nmatches with existing components of outfit; (2) Compatibility prediction:\npredicting the compatibility scores of given outfits. Experimental results\ndemonstrate the great superiority of our proposed method over others.","url_abs":"http://arxiv.org/abs/1902.08009v1","url_pdf":"http://arxiv.org/pdf/1902.08009v1.pdf","source":{"archive":"pwc-archive (Hugging Face), CC BY-SA 4.0","snapshot":"2025-07-28","licence_url":"https://creativecommons.org/licenses/by-sa/4.0/legalcode","row_kind":"abstracts"},"code_links":[{"paper_slug":"dressing-as-a-whole-outfit-compatibility","repo_url":"https://github.com/CRIPAC-DIG/NGNN","is_official":1,"mentioned_in_paper":1,"mentioned_in_github":1,"framework":"tf","reach":null}],"tasks":[{"task_slug":"recommendation-systems","task_name":"Recommendation Systems"}],"methods":[],"datasets_introduced":[],"methods_introduced":[],"results":[{"leaderboard":"/sota/recommendation-systems-on-polyvore","task":"Recommendation Systems","dataset":"Polyvore","model":"NGNN","rank_in_archive_order":3,"of":3,"metrics":{"Accuracy":"0.7813"},"uses_additional_data":false}],"syntology":{"syntology_url":"https://syntology.ai/paper/1902.08009","atlas_url":"https://app.syntology.ai/?focus=1902.08009","mcp":null,"developers":"https://syntology.ai/developers"},"arxiv_metadata":null,"syntology_extracted_results":null}