Could Algorithms Become Fashion Trendsetters?

Don't wear bytes after labor day

Machine Derived Shirts
Machine Derived Shirts
Stitch Fix

We have trained machines to dream, but can we make them follow their dreams of moving to New York to become a fashion designer? In a world where startups use algorithms to find the next fashion trends, it's not impossible that computers could go beyond finding what's popular, and start creating it. Stitch Fix, a fashion startup that aims to provide a personal shopping experience remotely, already uses machine learning to understand its customers' tastes. Last week, Stitch Fix data scientist TJ Torres poked into the potential future of computer-generated clothing designs.

Titled "Deep Style: Inferring the Unknown to Predict the Future of Fashion," Torres' post details a process not unlike that used by Google this summer to generate those really freaky images with dog faces everywhere. The core technology is an artificial neural network that can be trained to recognize a specific object by analyzing pictures of it, and gradually the computer builds up its own picture of what the object looks like. Sometimes, the computers' results are spot on. Sometimes, they misidentify Yoda from Star Wars as a giraffe. It's a kind of machine learning, and even the bad answers are informative.

For fashion, Torres took neural networks, and fed them pictures of clothing. In response, the network generated its own image. After being fed more images, the network's output eventually evolved away from just replicating the original shirts to arrive at new styles, patterns, and variations on the theme of shirts. The process is still in the early stages, so we can’t say that machine-derived patterns will sweep the next Fashion Week. But that’s a direction this could go. Says Torres:

Finding predictive signal in extracted stylistic concepts from images of clothing would represent a big leap forward in the modeling possibilities for our recommendation systems. More broadly, developing algorithms to quantify abstract concepts like style, fashion, and art may one day move us forward toward a more complex understanding of how we as people process and analyze abstract unstructured data. At Stitch Fix, we're in very early days of this research and have yet to implement it into our pipeline. Our ultimate goal is to employ this work to boost the combination of the recommender system and stylists to provide even better personalization than we see today.

When machines start learning the nuances of fashion, who will be the first to tell them they don’t have bodies to try it on?