AI coordination that has even entered the online mall’Looks like these pants on that jacket’

Artificial intelligence (AI) fashion coordinator function jointly developed by Pohang University with Samsung C&T. Samsung C&T Online Fashion Mall (www.ssfshop.com) recommends other fashion items that match when you select specific clothes. [사진 포항공대]

It’s good news for anyone who’s heard the word’failure’ and’fashion terror’. An artificial intelligence (AI) fashion coordinator has appeared that recommends matching pants, skirts, and shoes when choosing a jacket or top at an online shopping mall. It is the product of the online era and the development of artificial intelligence.

POSTECH (Pohang University) announced on the 8th that the research team of Professor Song Min-seok of the Department of Industrial Management Engineering developed an AI fashion curation service through joint research with Samsung C&T and provided it to the SSF Shop (www.ssfshop.com), an integrated online mall for the Samsung C&T fashion division. .

The developed AI learns the styling created by fashion experts and recommends clothes that best match the clothes selected by customers who visit the online shopping mall. Existing fashion AI only statistically processed and displayed clothes that customers bought together or found similar clothes. However, the AI ​​developed by POSTECH recommends’from head to toe’ in a style that matches bottoms, coats, shoes and bags when customers choose a top.

Researchers at the Institute of Electronics and Telecommunications (ETRI) are demonstrating the Self-Growing Multi-Intelligent Fashion HOW developed in October last year. [사진 전자통신연구원]

The research team also succeeded in localizing the dataset for fashion AI learning. The existing fashion AI had to rely on data from foreign companies such as’Polybore’ in the US and’Alibaba’ in China for learning. AI learned in this way was difficult to commercialize because it is far from Korea’s fashion trends. The research team created a matching dataset for fashion AI learning consisting of hundreds of thousands of product images and a million style sets.

The style set included in the dataset was constructed with the help of brand-specific fashion experts from Samsung C&T’s own brands such as ‘8 Seconds’,’Binfall’, and’Guho’, as well as brands in online stores. In addition, Samsung C&T’s self-developed visual intelligence AI engine’s clothing attribute extraction function and clothing object detection function were utilized to include various and rich information such as color, material, length, and location of clothing.

The learned AI recommends 20 matching style sets for hundreds of thousands of clothes sold at Samsung C&T every day. In the past, it took more than 30 hours to recommend a total of millions of styles, making daily service impossible. The POSTECH research team shortened this to be able to process it in real time.

Jarvis, an artificial intelligence assistant from the sci-fi movie Iron Man, is a multi-intelligent AI assistant who understands and executes all commands of the protagonist. [사진 영화 아이언맨 캡처]

“The AI ​​that has learned the specialized data unique to Samsung Fashion is now equipped with a considerable level of styling ability, so that customers can enjoy shopping conveniently with the help of their own AI stylist in the near future. The goal is to do it.”

Professor Song Min-seok said, “Currently, it recommends matching fashion items based on artificial intelligence, but through further research, the system is designed to produce clothes that match height, body shape, skin color, etc. We plan to develop it.”

Pohang University is not alone in developing online fashion coordination using artificial intelligence. Electronics and Telecommunications Research Institute (ETRI) also developed and unveiled’Fashion How’, an artificial intelligence coordinator that recommends clothes to wear according to the situation in October last year. Part of the task titled’Study on the source technology of self-growing complex artificial intelligence’, Fashion Howe recommends clothes that fit the time, place, and situation (TPO: Time, Place, Occasion).

Song Hwa-jeon, a senior researcher at ETRI’s Complex Intelligence Lab, said, “Language artificial intelligence, visual artificial intelligence, speech recognition, natural language processing, and other artificial intelligence technologies are used in combination to process human needs like Jarvis, the AI ​​assistant of the science fiction movie Iron Man. Haeju is aiming at developing complex artificial intelligence.”

Joonho Choi, Scientific and Future Reporter [email protected]