Microsoft is using dogs to demonstrate its machine learning chops. In the latest display of the power of deep neural networks, the Microsoft Garage incubation lab released an iPhone app and web app called “Fetch!,” which identifies the breed of a dog based on a picture. Like the previous apps and services in the series, it’s powered by Microsoft’s Project Oxford artificial intelligence platform.
“We wanted to bring artificial intelligence to the canine world,” said Mitch Goldberg, a Microsoft Research development director who leads the Cambridge, U.K-based team that developed the Fetch! apps and experience. “We wanted to show that object recognition is something anyone could understand and interact with.”
The breed-identification technology was first shown at a Microsoft Research event two years ago, and has evolved in the time since to a consumer-friendly app, with some extra features. In addition to identifying the breed of dog within a certain degree of certainty, for example, the app can take an “informed guess” as to what type of breed a person most resembles.
The app, called Fetch!, analyses photos of dogs using the iPhone’s camera or photo library and then attempts to determine the breed. “We wanted to bring artificial intelligence to the canine world,” Mitch Goldberg, a development director at Microsoft Research, said in a statement. “We wanted to show that object recognition is something anyone could understand and interact with.”