Personalized Outfit Compatibility Prediction based on Regional Attention

Personalized fashion collocation has recently attracted more research attention as it provides more clothing choices that conform to fashion aesthetics and users' preferences. The great challenge is combining the item's compatibility relationship and users' preferences. Actually, user...

Full description

Saved in:
Bibliographic Details
Published in2022 9th International Conference on Digital Home (ICDH) pp. 75 - 80
Main Authors Chen, Yilin, Zhou, Zhouyi, Lin, Ge, Chen, Xiaoyan, Su, Zhuo
Format Conference Proceeding
LanguageEnglish
Published IEEE 01.10.2022
Subjects
Data models
Encoding
Optimization
outfit compatibility
personalized fashion collocation
Predictive models
regional attention
Visualization
Online AccessGet full text

Cover

More Information
Summary:Personalized fashion collocation has recently attracted more research attention as it provides more clothing choices that conform to fashion aesthetics and users' preferences. The great challenge is combining the item's compatibility relationship and users' preferences. Actually, users may pay more attention to some regions, which are also crucial to determining item compatibility. Therefore, this paper takes regional attention as the fusion point, proposing a personalized outfit compatibility prediction model that consists of two modules. Firstly, the User Outfit Preference Encoding module encodes the item's visual features to obtain the user's initial preference for the given outfit. Then, the Regional Attention Compatibility Scoring module measures attention to item regions from user and outfit compatibility, fusing them to calculate the personalized outfit compatibility score. Finally, we improve the model's predictive ability by jointly optimizing the two modules. The experiments on two real datasets demonstrate that our model outperforms the state-of-the-art methods.
DOI:10.1109/ICDH57206.2022.00019