Amazon’s Latest Feature Is An AI That Designs Clothes By Looking At Pictures

Amazon’s Latest Feature Is An AI That Designs Clothes By Looking At Pictures
Design

The e-commerce giant is trying to take a new approach to fashion

Leo Lutero
  • 4 september 2017

What makes a collection sellable? Aspiring fashion retail giant Amazon is planning to answer that question with an AI designer. The company is building a neural network that will make clothing design an automated process, MIT Technology Review reports. For this AI designer, Amazon innovators from Lab 126 created a generative adversarial network or GAN. The GAN is an AI scheme developed by Google engineers that allows computers to understand complex human tasks, continually improve, and generate its eerily human output. In essence, GANs use two neural networks that compete against each other. Basically, one generates the output, the other verifies it. The generating-end improves itself to pass the verifier while the verifier works hard to be more discerning. Datasets, often vast, will help the technology tell the difference between what is stylish and the faux pas. The same system was recently used by Google which turned Street View data into convincing professional photography.

For data, the system has a lot to work with. In the era of self-chronicling, platforms like Instagram have become significant players in the fashion industry. By harvesting the images and using data such as likes and comments on certain looks, Amazon’s AI fashion designer can put a finger on what’s becoming on-trend and what’s ending up on sale bins.

In recent years, fashion retail has turned upon itself. From big designers getting exclusive power to create trends, the next big thing can be a user-generated movement. Edited, a big data analytics company is already helping fashion brands like Topshop and Desigual. The company processes millions of product images, providing their clients a clear idea of what styles are statistically more sellable.

While Amazon is still keeping the project under covers, it’s an interesting one to look out for. In 2016, the long process from the runway to stores have seen several attempts to shrink. The industry is slowly becoming a race of who can spot a trend first and who can respond to the demand. Amazon’s silicon fashion designer might just outrun everyone.

Amazon

What makes a collection sellable? Aspiring fashion retail giant Amazon is planning to answer that question with an AI designer. The company is building a neural network that will make clothing design an automated process, MIT Technology Review reports. For this AI designer, Amazon innovators from Lab 126 created a generative adversarial network or GAN. The GAN is an AI scheme developed by Google engineers that allows computers to understand complex human tasks, continually improve, and generate its eerily human output. In essence, GANs use two neural networks that compete against each other. Basically, one generates the output, the other verifies it. The generating-end improves itself to pass the verifier while the verifier works hard to be more discerning. Datasets, often vast, will help the technology tell the difference between what is stylish and the faux pas. The same system was recently used by Google which turned Street View data into convincing professional photography.

+AI
+AI
+amazon
+apparel
+data
+Design
+Fashion
+Google
+retail
+technology
+Virtual Commerce
+work

Learn About Our Membership Services

Need Research Help?
As a member you can ask us any research questions and get complimentary research assistance with a 4-day turnaround. Reports inclde stats, quotes, and best-inclass examples on research topics.
Remain Informed & Strategic
We publish several trends reports each month. By becoming a member you will have access to over 100 existing reports, plus a growing catalog of deep topical analysis and debrief-style reports so you always remain in the know.
See Trends Come To Life
Meet your peers and immerse yourself in monthly trend and innovation webinars and discounted conferences.
No search results found.