Resetting Expectations: Some Augmented Reality Links

1. Mobile Devices and AR: Besides employing the location of users (Wikitude), there are generally two ways to overlay data onto the real world: through markers ( (2D) bar codes) or through automatic object/image recognition algorithms (“markerless”). The Economist gives a good overview of the different mobile applications that are starting to emerge and lists a few areas where AR makes sense such as shopping (letting house-hunters which properties are for sale) and events (giving sports fans access to stats and player bios).

2. 5 Barriers to a Web That’s Everywhere: @gnat linked to a recent RWW post, that lists high-level challenges the AR industry needs to address, including spam and security, interoperability, user experience, and openess.

3. For technical challenges facing AR developers, I recently had a chance to visit with computer scientist and Everyscape CTO/founder, @mok_oh, who’s also been blogging about AR. In the first of two posts, he points out that accurate object and image recognition remain formidable technical hurdles (“accurate registration of the virtual objects with the real-world image”). Without object and image recognition, Mok points out that some of the more well-known AR apps may not actually be augmented reality apps in the classic sense (” … there’s not much difference between this and Google Maps on your mobile phone”).

In a follow-up post, Mok warns that too much hype may be worst thing that can happen to AR. Serious technical problems need to be resolved:

I still think we need to continue to expand/expound on vision algorithms (e.g. image tracking, image detection/recognition, etc.) and couple that with other sensors (e.g. Wifi, RFID, Bluetooth, accelerators, gyros, GPS, compasses, etc.) to more precisely tell people what they’re seeing in an interactive and augmented sense. The level of precision provided by current apps are good from a mapping perspective (i.e. the 2D “aerial” view), but not good enough from a first-person’s ground perspective.
… Perhaps, we need to reset people’s expectations somehow, or rebrand the words to something else. Because I really do think that there’s plenty of use for AR-inspired technologies as being defined by Layars and Wikitudes of the world.

Everything I read indicates that the more likely scenario in the near future is that AR applications will use a combination of sensors (like a GPS) and markers. In contrast, accurate markerless AR is a distant dream, that will remain locked away in the world of science fiction for years to come.

(†) Mok was at Foo camp last month and I had a chance to talk to him about AR and related topics. Given that he has long worked in the relevant fields within computer science, I take his word on the state-of-the-art in computer vision.

tags: , , ,
  • ohm

    What Mr. Oh is trying to say is he can’t do it. He failed with Everyscape as fact and this is simply a waste of your readers time. I don’t understand why people are so intrigued with a glorified video game producer who created some 3D models which at glance are “cool” yet hardly practical to consumers. He rails against the practical applications of Google Street simply because he has no choice. He gets face time simply as an alternative to the monopoly that is Google in certain realms of its umbrella. Educated and generally a kind soul I’m sure but resting his laurels on distant dreams is just solidifying that he can’t achieve his real ones.


    • ohm,

      At least Mok put his thoughts on the record in his blog. OTOH, you mounted unsubstantiated personal attacks without actually addressing the issues raised: object and image recognition in the AR setting are formidable (and unsolved) problems. Many experts will tell you that.


  • bowerbird

    you forgot to close your footnote tag.


  • ohm,

    I agree with Ben that AR in general is not a solved problem.

    It also seems like your problem with me isn’t solved either. ;-)

    Ben, wonderful post. Thanks.

  • Generic markerless AR is a distant dream indeed. However very generic AR applications are quite rare any way. So using faces, hands, or drawings(like we do) as markers is somewhat markerless AR.

  • Lennert Dorman

    But will users really interact with augmented reality on their mobile phones? Won’t it look really silly? Like we still don’t use videophones. I want augmented reality invisible on my glasses not on my phone where I’m the laughing stock of everybody.

  • It’s likely time to broaden the definition of AR. Next year it’s going to gobmainstream in layman terms as adding a layer graphics to real time video feeds. Markerless AR may need some definition work too.

  • Lennert — yes, that would be quite silly (and accident prone).

    Ron — agree that perhaps separating these two definitions might be worthwhile. Or even some blend of the two.

    While markers are meant to be a very specific mark, we can perhaps blend the marker/markerless definitions. Image detection/recognition tech is quite well defined now and being used in cameras (e.g. the cam recognizes faces and only takes the pic when the person smiles) to iPhoto facial recog feature (which works quite well).

    So perhaps certain things, such as faces, can be detected as “markers.” So in some sense, it’s a blend between marker and markerless techs, where certain category of things instead of a specific marker can be detected, and used for “AR” purposes.