The science behind Bing Vision 2D object recognition

Get Microsoft Silverlight

If you haven’t seen or tried Bing Vision, it’s pretty cool stuff. Even though iPhone users have had the privilege for more than six months now, Windows Phone Mango users (in some regions) will soon too be able to search book, CD and DVDs by just scanning its cover.

In a presentation about computer vision, Rick Szeliski from Microsoft Research’s Interactive Visual Media group talks a lot about the advancements in computer vision technologies MSR has made in the last decade including but not limited to Kinect. One other recent project of note was called “Lincoln”, by Larry Zitnick, which formed the basis for Bing Vision.

Since Rick tries his best to generalize the algorithms involved, I’ve embedded a clipping of the video above if anyone’s curious about the technology behind it all. Apparently it involves “kd-trees”, not magic.

One other interesting tidbit Rick mentions during the presentation is the fact that whilst recognizing people en mass is hard, it’s already viable for smaller groups like families. The fact that he uses the example of logging in a PC in light of the rumors that Windows 8 will support such a feature is at least note-worthy evidence of the feature’s practicality.

8 insightful thoughts

  1. and here we see why suggestions that Microsoft should sell Bing are silly – it synergises too well. But it does raise, in my mind, whether Microsoft should do more to commercialise this particular technology outside of “Bing Vision”

  2. >Even though iPhone users have had the privilege for more than six months now,

    Except if you live in Australia

  3. Anything to do with Bing I was never really interested in, up until recently when I have found out just what awesome innovation Bing can offer! Keep up the good work Microsoft! 😉

  4. Nice Lecture.
    Search and other online services, cut across ecosystem, platform and devices.
    Those guys who suggest MS to sell Bing, may not understand the business.
    Moreover it is crucial to have advance search technology. There are some many information around us, and not many of us organize the world, in a way that can be naturally understood by another person.
    Most of us have different perspective on things. To fully leverage our collective conscious, we need a way to obtain knowledge from it fast and efficiently. A good search engine would likely be the answer.

    Going back to the point of, search engine cutting across ecosystem. That means it can be some sort of foundations for platforms or ecosystem. MS is about platform, it would be crazy to let another company run your foundation or control your pillar. MS is too smart to let that happen.

    Thanks for reading.

Comments are closed.