[2F1] [General Session] 12. HI / Education Aid

[2F1-05] Proposal of fashion coordinate system based on color harmony

〇Takumi Saito¹, Mizuki Oka¹, Kazuhiko Kato¹ (1. University of Tsukuba, Graduate School of Systems and Information Engineering)

Keywords:fashion, color harmony, recommendation system

In recent years, with the development of fashion online shops or recommendation services, images of clothes can be easily collected. Recently, various models using machine learning or deep learning are proposed on the collected data. Moreover, various color theories have been proposed and shown that color combinations affect greatly on fashion.

In this research, we propose a system that recommends color combinations for clothes given an input image of a person. The proposed system is built by two models: the first is a parsing model that divides an image into regions by clothes; and the second is a harmonic evaluation model that evaluates the color combinations for clothes based on the color harmony theory.