New AI tool claims to 'change the landscape of online ads' by connecting shoppers to goods using images
Image: ZAC
Imagine that you are searching for a brown leather sandal online. You know what it should look like, but don't know how to describe it. You search "brown sandal" in Google, which serves up many results—but none of them are it.
Now, an AI platform called The Discover Machine, created by the startup Z Advanced Computing (ZAC), can use a photo you upload to deliver the result you want. The Discover Machine, available as a mobile and desktop app, uses image and pattern recognition to find the right image.
The primary users of the product, said ZAC, are consumers, who can find what they are looking for, merchants, who can display what's in their catalogs, and fashion web publishers and bloggers, who can earn revenue from the automatically generated visual ads.
ZAC offers two approaches for the platform. One approach is what they call an "image referral network."
"Something like searching for shoes turns out to be very difficult for traditional image recognition," said Dr. Bijan Tadayon, CEO of ZAC. Why? Text results produce generic results that typically aren't useful. And even Google's image search, Tadayon said, can produce things with similar shape and color, but are sometimes not even the right object.
ZAC, instead, allows customers to directly upload photos to the platform.
"The way we're approaching it is like how the human mind approaches it," said Tadayon. "A buyer can upload a photo on their smartphone and the analyzer will take the user right to the merchant's website with the item," he said. "It's good for merchants; it's good for buyers—they get exactly what they want."
Another, more indirect way it can connect customers to products is by sending the image to the merchant's website, which could pass it to ZAC for analysis, delivering the product that's located in that catalog (if it exists). "The platform would act like a smart visual search engine for the merchant to help the consumer find what they are looking for in its catalogs," said Tadayon.
SEE: AI's largest-scale innovations aren't happening in cars or robots but in customer service
Another application is called "image ad network." When someone looks at a fashion website or a blogger's article, for example, ZAC can give the image to the engine, which it will find on a website. The difference is that in this scenario, they will put the automated visual ad directly on the publisher's website.
ZAC says its targeted image-based ads is happening "on a large scale for the first time in the industry."
The team of developers at ZAC claim that their platform is a "breakthrough in AI and image recognition science," something that nobody else—including Google, Microsoft, Facebook, and others—can do.
ZAC's approach "is iterative, modularized and reusable," said Tadayon. "For example, once the platform learns how to detect a zipper in a shoe, it can be reused for zipper recognition in other objects/contexts, such as bags and clothing (e.g., pants and jackets)."
SEE: Artificial Intelligence and IT: The good, the bad and the scary
While the current technology is focused on recognizing shoe and clothing images, it's easy to imagine how the platform could be expanded to other products as well.
"This totally changes the landscape of online ads," said Tadayon.
The 3 big takeaways for TechRepublic readers
- ZAC claims to offer something unique by producing online ads generated through images, not text.
- The machine-learning platform can be applied to searches from shoppers that will lead to the product on a merchant's website, or to serve merchants by generating targeted visual ads based on a customer's browsing history.
- The intended users of the platform include shoppers, merchants, and bloggers or other publishers.
Also see...
Source: http://techrepublic.com.feedsportal.com/c/35463/f/670841/s/4ede41f4/sc/28/l/0L0Stechrepublic0N0Carticle0Cnew0Eai0Etool0Eclaims0Eto0Echange0Ethe0Elandscape0Eof0Eonline0Eads0Eby0Econnecting0Eshoppers0Eto0Egoods0Eusing0C0Tftag0FRSS56d97e7/story01.htm
