Geo-specific 3D map data will be AR’s “unsung hero” as we’ve written. Using a combination of object recognition, mapping, localization and inertial odometry, AR can fulfill its promise.
Of course that’s a complicated set of tasks, and we’re not quite there yet. There’s a difference between today’s AR’s execution, and ARkit videos about navigating grocery stores and such. That vision requires product data, object recognition and geo-relevant 3D mapping among other things.
“3D map data is the scaffolding for the 21st century,” said Scape Co-Founder and CEO Edward Miller at the recent AWE Europe (video below).
Miller explains how humans have always tried to render the physical world through representative media. Another word for that is maps. In fact, the roots of the word “geography” are geo and graph… as in representing the physical earth’s layout through a 2D grid.
But one thing most maps have in common is their designated purpose to be understood by humans. As we enter the world of AR (and self driving cars for that matter), geographic data will need to be seen and understood by machines, also known as computer vision.
This leads to object recognition which is the “understanding” part. An AR device’s ability to understand its surroundings is enhanced by geo-relevant 3D mapping data that it can essentially match to what it’s seeing. This is one of the things separating true AR from heads up displays.
“We’ve got to have a one-to-one map of the world so we can precisely align 3D content with the physical world,” said Miller. “Because if we don’t know precisely where we are, the digital and physical world won’t align.”
This is a key part of what SuperVentures partner and AWE Conference lead Ori Inbar calls the AR cloud. Having that mapping data available in the cloud can help AR devices map the areas they enter (the M in SLAM), rather than exhaust compute muscle on already-chartered territory.
Beyond processing efficiency, it’s a matter of effectively matching anything a device sees to a database of 3D images. That enables any AR device to be a lot more powerful. Think of it sort of like marker-based AR, but for the world around us instead of a few recognized QR codes.
“We don’t want stuff like this where you’re augmenting cereal boxes or chocolate bars,” said Miller. “We want rich digital experiences overlaid onto the physical world. and that requires data… a huge amount of data.”
That’s the short and simplified version. For a more in depth narrative, check out Miller’s full presentation below.
Disclosure: ARtillry has no financial stake in the companies mentioned in this post, nor received payment for its production. Disclosure and ethics policy can be seen here.