Although Japan’s declining and aging population has done little to stimulate the domestic fashion industry, AI is getting very positive reviews from the country’s still massive JP¥9 trillion (US$83 billion) apparel market.
In Japan as elsewhere the performance of fashion shopping websites and other online clothing shopping platforms is expected to continue to progress. AI is already boosting Japanese consumer convenience and reducing tedious labour for workers. Improving online shopping conversion rates is a key new area for AI in fashion, while other emerging AI roles include trend prediction and personalized style recommendations. And judging cuteness.
Fashion Trend Forecasting
Many Asian and global fashion trends are rooted in the streets of Shibuya and Harajuku. What people are wearing in trendy neighbourhoods is very valuable data. Tokyo-based Fashion Pocket uses AI to classify and evaluate street fashion images in Japanese and global media and identify trends such as colours and patterns and fashion sense and other qualitative information. Fashion Pocket also provides image analysis technology based AI system services to large apparel companies.
The Fashion Pocket AI generates analysis reports that sum up fashion trends extracted from tens of millions of images and other data every day from all over the world. The company expects the service to expand across the fashion industry in the future to buyers, designers, researchers, clothing stores, clothing schools and clothing companies. It’s thought the AI can also help improve retail turnover rates to reduce inventory and waste.
In addition to objective global trend forecasting from big data, fashion monitoring and prediction services could also for example find the Parisian look for this autumn, or show designers what’s currently in vogue in Seoul nightclubs.
Clothing Style Recommendation
Another scenario with huge potential is clothing style recommendation, increasingly used by companies to deliver targeted product suggestions to customers. Online clothing platform D Fashion uses an AI provided by Japanese mobile giant DoCoMo to turn user photos into sales pitches.
The company offers a service that uses AI to search for products similar to those in user-submitted images. The AI can not only recommend similar products from photos taken on smartphones, it also automatically identifies and stores the items it sees in categories such as tops, hats, pants, etc.
AI Cuteness Evaluation System
Few words better summarize Japanese youth culture and fashion then “kawaii” (cute). Whether in manga, anime, or on Takeshita Dori, girls in particular present with obvious attention to loveliness. Tokyo-based Insight Lab and Hokkaido University have developed an AI system that rates the kawaii factor of clothing.
After analyzing users’ uploaded clothing images, the newly developed AI quantifies the images according to what the company calls “emotional components” such as “cute”, “girl”, “female”, “date” and “nature”. For example pleated dresses with floral patterns score 98 percent on “cute” and 84 percent on “girl”; while adult knitted dresses score 77 and 45 percent respectively on these factors.
There are about 150 emotional components — others include “summer”, “easy to wear”, “party” and “I want to post on Instagram.”
Future Outlook and Trends
AI is becoming très chic in the Japanese fashion business. It has developed the ability to predict trends and is helping sellers decide which designs to produce. We can expect further breakthroughs in image retrieval and recommendation systems for style and ensemble matching, fashion trend prediction, etc.
Data sources for AI analysis include store cameras, customer history purchase records, images uploaded by customers themselves which reflect personal preferences, lots of image data from social media sites, and fashion and trade show and other video data. This is the fuel that is driving AI in fashion in Japan.
Author: Yuu Rirou | Editor: Michael Sarazen