Pioneering study takes a unique approach to applying cutting edge deep learning technology to apparel design

 

TAIPEI, TAIWAN - Media OutReach - February 13, 2018 - Appier, a leading artificial intelligence (AI) company, demonstrated pioneering research that showcased AI's ability to be creative at the 32nd AAAI Conference on Artificial Intelligence (AAAI-18), one of the world's leading conferences in AI.

 

AI's ability to create and design items that humans will find appealing is considered the next important breakthrough in the field - and Appier's research suggests that the day when AI can help designers in their day-to-day work might be closer than we thought.

 

Appier demonstrates AI's ability to recommend matching clothes and create beautiful apparel designs at AAAI 2018

 

The Appier team, comprising Yong-Siang Shih, Dr. Kai-Yueh Chang and Dr. Hsuan-Tien Lin from Appier and Prof. Min Sun from National Tsing Hua University, designed an AI model that first mimics the human ability to put together and design outfits, after observing and learning from examples of good design.

 

They did this with deep learning techniques in combination with a new Projected Compatibility Distance (PCD) methodology which they designed, that discerns the distance between compatible items.

 

This compatibility distance is then used to run through a unique application of Generative Adversarial Networks (GANs) to generate novel compatible apparel items when apparel images are input. The output results capture the designers' concept of how different apparel items combine with one another and is then used to create new apparel designs.

 

"This is an exciting step forward towards realizing creativity in AI. Appier's vision has always been for AI to help humans in all aspects of life, and this includes creativity. We don't however, see AI replacing human creativity, just amplifying it," said Chih-Han Yu, co-founder and CEO, Appier. "At Appier, we are always striving to close the gap between academic AI and the business world and this pioneering work by our employees is one example of the various innovation projects that we've adopted to achieve this."

 

"The team chose apparel design to apply our model because of the inherent complexity and subjectivity in matching different outfits as well as the potential for application in the business world. We also wanted to show the potential for creativity in AI using our unique applications of deep learning techniques in compatibility distance and GAN. The designs generated by our model easily passed the human test which has given us tremendous encouragement," Dr. Hsuan-Tien Lin, Chief Data Scientist, Appier.

 

The paper, entitled "Compatibility Family Learning for Item Recommendation and Generation", can be viewed in its entirety here. (riauone/MOR).