|
Briefing
|
- Fashion Recommendation AI – Researchers from University of Texas at Austin, Cornell Tech, Georgia Tech and Facebook AI Research introduced Fashion ++ artificial intelligence system that recommends more fashionable changes to outfits
- AI Model Training – Trained image generation system with over 15,000 fashion images, plus 12,000 outfits and their alterations as negative examples
- What It Does – Provides string of edited images from least changed to most fashionable to give user ideas how to tweak or improve outfit
- Survey Results – 92% of almost 300 respondents recruited from Amazon’s Mechanical Turk thought Fashion++ recommendations were more fashionable, while on outfits already considered fashionable, 84% said AI’s suggestions were similarly or more fashionable
- Next Steps – Train AI with more images from broader set of sources, such as social media platforms like Instagram, and adding recommendations according to individual style, occasion or body shape
|
|
Accelerator
|
|
|
Business Model and Practices
Business Model
and Practices
|
|
|
Sector
|
Consumer Non-Durables, Information Technology, Wholesale and Retail Trade
|
|
Function
|
Research and Development
|
|
Organization
|
Cornell Tech, Facebook Inc., Georgia Institute of Technology, University of Texas at Austin
|
|
Source
|
|
|
Original Publication Date
|
October 23, 2019
|
