Mobile augmented reality developer Blippar's mission is to construct a computer vision map the world, with visual recognition of thousands of notable buildings, bridges, castles, holy places. A major step toward that goal has been taken by the company this month with the announcement of its landmark recognition API.
The API, available now via the Blippar app for iOS and Android, is capable of recognizing famous structures like the Eiffel Tower or the Statue of Liberty based on computer vision—and without the assistance of GPS. Once identified, the app delivers more information about the landmark from sources like Wikipedia or Blipparshere, Blippar's own knowledge graph.
"On the heels of the AR City app, an industry breakthrough in machine learning entire cities with the potential to disrupt current maps in the navigation and tourism sectors, as well as Urban Visual Positioning, a disruption in location-based AR that leverages computer vision for better localization accuracy than GPS alone, Blippar is continuing to change the way people interact with and experience their surroundings," said Ambarish Mitra, Blippar CEO and co-founder, in a statement. "This is a great way to satisfy our curiosity and discover more about the world around us."
Blippar claims to deliver 91.6% accuracy in open set testing, with support for 2,164 landmarks at present. I tested it out with some less-iconic angles from community-generated 360-degree images on Google Maps for the Taj Mahal and the Golden Gate Bridge, and Blippar quickly returned the correct results.
But what about more obscure landmarks, such as the statue of Vulcan in Birmingham, Alabama? Blippar was able to recognize the largest cast-iron posterior in the world almost instantaneously. Sticking with statues in Alabama, the app was also able to identify a wide view of the Boll Weevil Monument in Enterprise, Alabama.
I was able to force one false-reading, but it was still in the ballpark. The app mistook a photo of Arulmihu Sivasubramaniya Swamy Kovil in Colombo, Sri Lanka, for the Kapaleeshwarar Temple in Mylapore, Chennai and the Sri Maramariamman Temple in Kuala Lumpur. But at least it had the right idea, as they are all Hindu temples.
As impressive as its latest feat is, Blippar faces a looming challenge for relevance, with Apple adding image recognition features to ARKit and Google launching its own computer vision capabilities. Neverthetheless, computer vision capabilities in all forms will continue to be useful as developers look to create smarter AR experiences.