Scientists develop AI system that can ‘design clothes’

Could this be a stepping stone to robot fashion designers in the near future?
Could this be a stepping stone to robot fashion designers in the near future?

Scientists have developed a system that can design new clothes based on people’s personal taste using artificial intelligence (AI).

Online retailers are already using AI to recommend products to buyers, but researchers wanted to take it a step further by creating an algorithm to “produce new clothing designs”.

The project is an attempt to test whether machine learning can help the fashion industry as well as consumers, particularly online shoppers.

The team trained an algorithm known as the Siamese convolutional neural network (Siamese-CNN) to learn and classify a user’s preference for certain items.

Clothing store
Will we be wearing clothes designed by AI in the future? (Mike Egerton/PA)

Siamese-CNNs are generally used to find similarities or relationships between two comparable items. For instance, if a user indicates a preference for a blue plaid shirt, the system can recommend similar items of clothing, such as a plain blue shirt or a plaid scarf.

Using Siamese-CNNs as the base, the researchers then developed another neural network called a generative adversarial network (GAN) to generate realistic images of “novel fashion items that maximise users’ preferences”.

The researchers say their system can suggest items to buy from existing designs, modify existing designs and generate new designs tailored to a specific individual’s preferences.

Study leader Wang-Cheng Kang, from the University of California, San Diego, said: “Building effective recommender systems for domains like fashion is challenging due to the high level of subjectivity and the semantic complexity of the features involved.

Personalised fashion clothing generated by AI.
The images on the left are real, while those on the right have been generated by AI tools – the values are preference scores for each image (University of California, San Diego)

“We show that our model can be used generatively, ie, given a user and a product category, we can generate new images (in this case clothing items) that are most consistent with the user’s personal taste.

“This represents a first step towards building systems that go beyond recommending existing items from a product corpus, to suggesting styles and helping to design new products.”

E-commerce giants Amazon and Alibaba are said to be working on GANs and other AI tools, although the researchers say the use of AI in the fashion industry is still in its infancy.

The findings are published on ArXiv.