Tech in Action: Amazon’s fashion-forward play

The online retail giant taps into machine learning's potential for clothing design.

An astute fashion fan can scan Instagram, magazines and influencer videos to piece together what’s hot and spot emerging trends. Amazon, however, is working on taking the human out of that equation with a number of machine learning projects designed to produce software that spots the next hot trend destined for the runways.

As first reported by MIT Technology Review, the online retail giant showed off several products in development that use generative adversarial networks or GANs. They’re basically neural networks that go head-to-head when picking over data, challenging each other’s findings to get to underlying facts. One system produces an output that is verified by the other’s analysis of the same data. Both Facebook and Google are reportedly using GANs in product development.

When it comes to Amazon’s fashion tech, this means those networks scan photos of clothes in a certain style and then design brand new items that fit that style. Another team of developers looks at image metadata to determine whether a given look is in-style.

This could have obvious benefits to partnered retailers selling certain styles of clothes, or shoppers looking to stay up on trends. But perhaps the biggest beneficiary will be Amazon itself as it continues to expand its own clothing business.