Abstract
According to the FashionUnited.com, the global fashion and apparel industry is valued of three trillion dollars, making up two percent of the world's gross domestic product, which demonstrates people's great demand of clothes. Indeed, clothing plays a pivotal role in people's daily life, as a proper outfit (e.g., a top with a bottom) can greatly empower one's favorable impression. As each outfit usually involves multiple complementary items, like tops, bottoms, shoes and accessories, the key to a proper outfit lies in the harmonious clothing matching to a great extent. However, not everyone is a natural-born fashion stylist, which makes choosing the matching clothes a tedious and even annoying daily routine. It thus deserves our attention to develop an effective clothing matching scheme to help people figure out the suitable match for a given item and make a proper outfit. Thanks to the proliferation of online fashion-oriented communities such as IQON and Chictopia, as well as the e-commerce sites such as Amazon and eBay, where a tremendous amount of rich real-world data regarding users' shopping, reviewing and coordinating behaviors on fashion items have been accumulated, the researchers are hence enabled to investigate the code in clothing matching.
In a sense, the problem of clothing matching we pose here primarily requires modeling the human notion of the compatibility between fashion items. Despite of its significant value, compatibility modeling for clothing matching, which involves not only a large number of feature variables but also complicated factors such as the domain knowledge and user subjectivity, still suffers from the following tough challenges:
a) Absence of comprehensive benchmark. How can we construct comprehensive datasets to facilitate the validation of compatibility models?
b) Comprehensive modeling. Existing works mainly focused on measuring the compatibility based on visual images of items but failed to take their contextual information into account. Accordingly, comprehensively compatibility modeling with the multi-modal feature variables is largely untapped.
c) Knowledge incorporation. As an integral part of people's daily life, clothing matching domain has accumulated valuable knowledge. How to utilize such knowledge to guide machine learning is another challenge we are facing.
d) Interpretabilities. A realistic compatibility modeling scheme should not only give the final decision on whether the given fashion items are compatible or not, but also provide the underlying reasons. How to enhance the interpretability of the compatibility modeling?
e) Subjective Aesthetics. The aesthetics can be rather subjective, as different people may have different fashion tastes. Thereby, personalized compatibility modeling should be considered. Besides, several other challenges have been raised, for example, the confidence of various knowledge rules for clothing matching and multi-modal cooperative fusion of fashion items.
In fact, the huge economic value of the fashion industry has drawn great attentions from many researchers. Existing efforts mainly focus on tackling the problem of clothing retrieval, clothing recommendation, and fashionability prediction, while as far as we know, few of them have been dedicated to the compatibility modeling, not to mention the knowledge-guided and personalized compatibility modeling. Noticed this timely opportunity, and in this book we present some state-of-the-art theories on compatibility modeling. In particular, we first brief the way of building several comprehensive benchmark datasets. They have been released to facilitate other researchers. We then introduce how to model the compatibility which can be affected by complicated factors in a pure data-driven manner. To further boost the performance, we next present a novel knowledge-guided compatibility modeling framework, where two knowledge encoding methods are proposed. Thereafter, we enhance the interpretability of the compatibility modeling scheme from the two perspectives: prototype-wise and region-wise. Following that, we extend the general compatibility modeling to a personalized one. Finally, we conclude the book and figure out the future research directions in compatibility modeling towards clothing matching, including the compatibility modeling for outfits with multiple items.
This is the preliminary research on compatibility modeling towards clothing matching, and we expect it can evoke active researchers to work on this exciting area. If I have seen further it is by standing on the shoulders of giants.
Copyright (C) <2018> Shandong University
This program is licensed under the GNU General Public License 3.0 (https://www.gnu.org/licenses/gpl-3.0.html). Any derivative work obtained under this license must be licensed under the GNU General Public License as published by the Free Software Foundation, either Version 3 of the License, or (at your option) any later version, if this derivative work is distributed to a third party.
The copyright for the program is owned by Shandong University. For commercial projects that require the ability to distribute the code of this program as part of a program that cannot be distributed under the GNU General Public License, please contact <sxmustc@gmail.com> to purchase a commercial license.