DoubleMe compilationAt the most recent Silicon Valley VR meetup, Albert Kim, CEO of DoubleMe, demonstrated his company’s technology for creating realtime 3D models from a synchronized collection of 2D images. Subjects step into a small studio consisting of blue-screen walls and 8 inexpensive cameras. Capture computers take the synchronized video feeds and run them through a series of imaging algorithms to create a 3D model in realtime. What differentiates DoubleMe from other solutions is that it’s capturing motion, and it’s doing it using cameras instead of motion capture gear.

The results are pretty impressive, as Albert demonstrated for me:

DoubleMe is opening a San Francisco studio soon that will allow for free 3D modeling of you, your kids, or your pets. You can even create a 3D printout of yourself based on the model, presumably to use as an action figure. For more information, see DoubleMe.me.

This article may contain affiliate links. If you click an affiliate link and buy a product we may receive a small commission which helps support the publication. See here for more information.

  • I think it’s the video part that is impressive. And the fact they’re using low cost cameras, in comparison to lets say eight Kinect v2. The resolution leaves a bit to be wanted though, this web demo doesn’t look very sharp, and the left hand is duplicated. I wonder if they are experimenting with better cameras, or if that makes the calculations too heavy.

    Oh, and their domain is funky… a Double Meme it appears.

    • hkalbert

      Andreas,

      We’ve been working on Unity 3D integration on DoubleMe (then Voxelogram). The demo is simple static rendering of DoubleMe contents (in FVV format). As we re-started with DoubleMe, we are now moving into WebGL instead of Unity.

      For the domain, as I mentioned on my latest comment, we are re-designing DoubleMe web site. But until we finish it, I want to re-use the existing Voxelogram’s web site for the time being. :)

      If you have any question or suggestions, please e-mail me directly to kim@doubleme.me

  • Alkapwn

    That looks very impressive indeed. Maybe the holodeck isn’t as far off as we think. If you combine this with that path remapping technology you could have something amazing.

  • vrdragon

    Using multiple cameras to record a moving 3d image- is a very basic VR technology. Most VR beginners can do it. This is an absolutely commonplace way of creating 3D motion that has been around and available to anyone with basic set of skills for years.

    • Do you know of other companies or specific software I can research?

  • vrdragon

    I have been looking into theory research on this tech for a few years. I cannot suggest any practical apps. In English maybe you might like http://www.image.esys.tsukuba.ac.jp/~kitahara/data/pubs/2003_MFI2003_kitahara.pdf
    http://www.google.com/patents/WO2013112749A1?cl=en
    http://www.hindawi.com/journals/ijdmb/2010/247108/
    I agree with you in principal this kind of recording tech is good. But with this particulat one, I tried to access the link and it redirects to some unrelated company site not updated for a few years, feels a bit fishy and unclear where they got this tech from.
    Cheers

    • Thank you for the links. I haven’t made a study of creating 3D models using several multiple 2D images, so I thought it was pretty interesting, especially when the display flipped over to wireframe. I also downplayed the augmented reality part of this since we’re a VR site, but I was impressed that the model was able to stay fixed on the marker as the camera panned around. Sony was demoing something similar at GDC that me and Ben recorded (which also included dynamic shadows, which was awesome), but the video was never published.

  • umblous3

    Tried to download the software to play around but it was a bad download and the mail to their helpdesk bounced.
    Looks like scam, the website belongs to different company (voxelo) and mentions some Zenitum as original developer.?? When I searched for Zenitum, the guy(David) replied that they went into administration and was neither happy nor interested to talk me.

    • @umblous3 Hey, the voxelo/doubleme thing is addressed on the front page of their web site, so that doesn’t seem fishy. Don’t know about the Zenitum connection. The intention of the article was to show what was being demoed at SVVR; I didn’t devote any time toward investigating the company, since they’re not selling anything and I’m not making a recommendation. Just wanted to bring it up on your radar. Do let me know if you find out anything else though.

    • hkalbert

      Hi all,

      Thanks for all the interest!! I was quite impressed by all the comments.

      Let me give you a short history of DoubleMe. Original technology of DoubleMe was started as a pet project in Zenitum (I previoulsy founded in 2004) back in 2007. Then, I spun-off the project as Voxelogram, Inc. in 2009 as I see the actual working prototype. By 2011, we started to see the outcomes as you see the demos. It’s been more than 5 years of private R&D.

      Now I saw the new opportunities in VR spaces with our 3D reconstruction technology. I rencelty decided to create DoubleMe with fresh aim, a free & walk-in studios to generate and share personal 3D characters without any professional help

      At the moment, we are designing new DoubleMe.me web site. For the time being you will see the remnants of either Voxelogram or Zenitum. If you have any question regarding DoubleMe, please email me directly at kim@doubleme.me

      The last commet on technology, the technical term for DoubleMe’s magic is called “Video based rendering.” Many already saw 3D rendition from multiple images, “image based rendering.” When the target is NOT moving or changing (rigid body), 3D extraction is not that difficult. Someone already mentioned a “basic VR technology.”

      However, when target model is freely moving, then rendering software has to consider complex operations such as occlusion problems. So called, “Non-Rigid Body” 3D reconstruction is not simple as we see from a simple “image based rendering.”

      If you are interested in playing with our FVVlib, please let me know. I took down the link recently but I can e-mail you the library so that you can build your own FVV file player!

      Contact : Albert Kim
      Mail: kim@doubleme.me

  • umblous3

    Mate would double check on this if I were you