One year ago, the Madison startup Markable rolled out its signature product: an app that could take photos of dresses, shirts, handbags or heels and tell you where to buy the clothes in the picture.
Today, that initial game plan has been scrapped. Instead of using its sophisticated artificial intelligence for a consumer fashion app, Markable is now selling its technology to online fashion retailers.
"People don't like to download apps anymore," said Joy Tang, Markable's CEO. "Their phone is so full of apps already."
The original Markable app was a one-stop shop for fashionistas. A consumer could upload a photo of a clutch handbag, or a model with a snazzy ensemble, and the app would identify the products in the image and highlight similar clothes available for purchase on retail websites.
That app is now gone. Starting this month, Markable’s tech will be found on the website AKIRA, a Midwestern fashion line based in Chicago. When visitors search AKIRA’s inventory, they’ll have a “camera search” option, where they can upload an image of a piece of clothing they like and find similar items to shop through in return.
Then, there’s what Tang calls the “reverse-engineered” version of the Markable technology. When a visitor clicks on a piece of clothing on AKIRA’s shop, they’ll be able to see if it has ever been modeled by a celebrity or fashion blogger. Click on a shirt, and you may see a photo of when that same shirt was previously worn by Taylor Swift. Shoppers then have the option of “completing the look” by buying the rest of T-Swift's ensemble.
Markable also offers “visual search engine optimization” to retailers. In other words, the software automatically creates descriptions for clothing that will make the items more likely to pop up during a Google search.
AKIRA is just the beginning, said Tang. Markable is currently in talks with five other retailers. The goal, she said, is to become the industry standard for online fashion shopping.
The software Markable has developed is no small feat: When it comes to image recognition technology, clothes are among the toughest things for computers to parse. Fabric can be twisted or contorted into all kinds of shapes or patterns, making it difficult for AI to figure out the patterns.
Tang said that makes fashion one of the next frontiers for image recognition technology.
"If you can do fashion, you can do anything else,” she said.
Tang said that the company’s algorithms have come a long way in the past year, especially since they added more computer scientists to their team to enhance the software’s capabilities.
“When we launched the app last time, we didn’t have our four PhD scientists,” said Tang. “The results were not that amazing."
The technology is on a different level now, said Tang. Before, the software took 10 to 20 seconds to analyze an image. Now, it can do so nearly instantly.
The image recognition tool is available to try out as a demo on the company’s website.