New image recognition app brings context to smartphone photos

New image recognition app brings context to smartphone photos

Deep learning will create a 'visual internet' where everyday objects will be recognized by smartphones, says an app developer.


We've all seen the QR code. That's the pixelated-like mobile phone-readable barcode that marketers place on or near products. Point your smartphone at the encoded icon, scan it with an associated mobile app, and the device links to a website.

Then there's the regular, stripy barcode that you can use to compare prices at brick-and-mortar stores.

Both of these technologies are likely to be superseded.

Image recognition

The Blip is a proprietary marker-less image recognition technology. It doesn't use any kind of barcode because it doesn't need to. The technology actually identifies the image and then delivers the marketing message.

I was able to successfully scan a Heinz tomato ketchup bottle and obtain smartphone-delivered ketchup recipes in a kind of virtual reality-like superimposition.

It's still early days for the technology, and although I was able to learn all about the wonderful things possible with ketchup, I wasn't able to obtain the same results with a can of corned beef hash—nothing happened.

The reason: Heinz has got a deal with the Blippar, the app's publisher, where it has developed a Blipping campaign so the bottle is known by the database.

Everyday object database

However, Blippar wants to take it further and get the engine to recognize unknown objects.

The company has loftier goals than just tomato ketchup recipes—however tasty—and wants to develop an image database of everyday objects.

"Point your phone at a dog and it will tell you what breed it is," Tracey Lien writes of the visual technology for the Los Angeles Times.

"Blippar wants to making 'blipping'—the act of scanning an object through the app—as habitual as opening Google Maps in a search for directions or opening Yelp for restaurant reviews," Lien writes.

Neural engine

A deep learning neural engine is the background science. Roughly, the computer learns what the object is from seeing repetitious, multiple images. It's a kind of elaborate pattern recognition.

As more advanced deep learning occurs—mimicking human thought processes, like categorization—the more accurate this kind of technology can be.

Ultimately, it's the kind of thing we'll be seeing in self-driving cars, for example. The technology learns the difference between roadside objects that appear similar. "Is that a person at the crosswalk or a lamppost?" Lien uses as an example.

'Visual Internet'

That's the idea. But it's not all there yet—Blippar can't identify objects arbitrarily.

"Right now, when you start the Blippar app, it can't tell you about the chair in front of you, but it'll recognize a bottle of Coke," Ambarish Mitra, Blippar co-founder and CEO, said in an interview with NDTV Gadgets. The Coca-Cola can is in its catalog.

However, NDTV Gadgets says "in the next five to six months, Blippar will be launching a 'visual Internet,' wherein the Blippar app will be able to identify objects that aren't necessarily in its catalog."

Blippar says it will use deep learning to achieve this.

Visual browser

"Our aim is for Blippar to become the visual browser of the future," the company says on its website.

"To achieve this, we aim to be the default Augmented Reality lens through which people unlock interactive experiences. And we plan for 'blipping' to become the eponymous verb that describes doing it," it says.

Follow Us

Join the newsletter!


Sign up to gain exclusive access to email subscriptions, event invitations, competitions, giveaways, and much more.

Membership is free, and your security and privacy remain protected. View our privacy policy before signing up.

Error: Please check your email address.


Show Comments