A group consisting of Waseda University and ZOZO Research Institute has developed a new AI technology that automatically interprets fashion images, which tend to be explained in ambiguous terms, and answers questions from users.
Fashion is characterized by different evaluations and images depending on people's tastes, values, and cultural backgrounds, and the expressions used by users are extremely vague, such as "casual", "formal", and "cute".Therefore, we answer abstract questions from users such as "How casual is this outfit?" "What makes this outfit casual?" Doing so is not easy, even for experts.
Therefore, this group is able to obtain answers to user questions by automatically learning the relationship between vague and diverse expressions (tag information attached to images) in coordinated images and related fashion with AI. Developed new technology "Fashion Intelligence System".Using this technology, for example, clothes tagged with "office casual" can be sorted in order of "office casual", or a user's clothes can be searched for "slightly casual" clothes. canIn addition, it is possible to visually indicate "which part is casual" in the retrieved coordinated image.
In this way, AI that automatically interprets and suggests fashion is expected to reduce the ambiguity peculiar to fashion and become a system that supports the selection of clothes to wear and items to purchase.
In addition, this technology can be applied to other fields as long as there are images and sentences and word information associated with the images, so vague descriptions such as architecture, art, furniture, and cooking are avoided. Further expansion into many other fields is also expected.
Paper information:【Expert Systems with Applications】Fashion Intelligence System: An Outfit Interpretation Utilizing Images and Rich Abstract Tags