Imagine aiming a smartphone or a tablet at a cereal box and, instead of seeing the static image printed on the box, you see an animated feature appear as if it's playing on the front of the box. The same technology could be used to turn a picture in a printed newspaper into a video or--if pointed toward a product--it could launch a game featuring animated characters interacting with that product.
That's the aim of a new technology being introduced by Autonomy, the London-based company best known for its enterprise software.
Aurasma, which is a core technology designed to power apps for the iPhone, Android, and other smartphone and tablet devices, is capable of recognizing images and objects and enhancing them not just by replacing them with video but by allowing virtual objects to interact with real ones, such as a cartoon character walking through he door of a real building.
A prototype of the software is remarkably easy to use and surprisingly robust. During a preview event in San Francisco, journalists were given the opportunity to aim an iPad 2 camera at a newspaper, a cereal box, a Harry Potter poster, and a picture of the Mona Lisa, and then watch as the images transformed from static to moving. I expected the demonstration to work with the reproduction of the Mona Lisa supplied by the company, but was surprised to get the same result when I pointed the iPad toward my Android phone after downloading a different and much smaller rendering of the same image.
Autonomy will license the technology to publishers, consumer product makers, and others who want to use it to "augment reality." A print newspaper, for example, could use the technology to enhance its advertisements so that readers could point their devices at the paper to get a richer experience than they could by simply looking at an ad. That's sometimes done now by printing bar codes (or QR codes) in the ad, but this technology would simply recognize the image itself.
Aurasma could also be used by product manufacturers to provide additional information. Visitors to a car show, for example, could point their phone toward the grill of any car to get a video or perhaps a game that lets you virtually drive the car. A travel magazine could not only bring its pages to life but could offer virtual tours of cities triggered when people point their phone toward a building. Someone could write an app that, when pointed at the Lincoln Memorial in Washington D.C., could get the late president to get up from his chair and give a speech.
The software does require that someone code information about objects. If you point it at a house, it might know it's a house, but it won't know whose house it is unless that information has been entered. It can, however, read text.
Autonomy CEO Mike Lynch said that while it can be configured to recognize pictures of certain people, it's not that good at recognizing actual people or animals. He said that if you point it at a dog, it might recognize that it's a dog but would have a hard time figuring out the breed. "It's designed for man-made objects, rather than things of nature," he said.
The technology is based on what Lynch refers to Bayesian inference, named for an 18th-century English mathematician and minister. "People like to think that the way in which you would solve these problems is to take lots measurements of all these things and puts lots of rules into a computer and it would then execute these rules and make decisions," Lynch said. "But it turns out that it's very hard to make that work." Instead, the algorithms behind Aurasma "are based on probability theory," which he says "[harkens] back to an amazing piece of work by an English country vicar called Thomas Bayes."
For more about Aurasma, listen to my 12-minute podcast interview with Autonomy CEO Mike Lynch.