Motion capture gets slightly less embarrassing

It looks like the profession of motion capture artist could soon become a slightly more dignified one, with one upstart company doing away with the trademark spandex suit and ping-pong balls in favor of a come as you are approach. That bit of of freewheelin' technology comes to us from Organic Motion, who showed off their wares at the big Games Developers Conference in San Francisco. According to the company, instead of relying on reflections from a few markers placed on a person's body, its system automatically visualizes and captures thousands of natural points on a subject, turning it into a complete 3D scan in real-time. On the downside, the current implementation of the technology (set to launch later this year) can only scan one person at a time, although a version that scans two people is apparently on track for 2008. On a related front, a team from Germany's Max Planck Institute for Computer Science have developed a system that can scan a high-resolution 3D image of a person and quickly turn it around into a model ready to be animated or, for instance, imported as an avatar into a game. It appears that it's slightly further down the line, however, with no indication given as to when, if ever, we might see it commercialized.

Read - Newsweek - Level Up, "Organic Motion's Marker-Less Motion Capture"
Read - New Scientist, "Animation tool puts you in the frame, or the game"

[TIMMAH! @ Mar 12th 2007 6:32PM]

Doesn't the ping pong ball approach place the markers at the joints in the limbs so that you can do tweening that looks reasonably natural? Seems like this approach might cause joint positions to "move" which might lead to weird results with things bending or flexing in places where they shouldn't.

[Twist @ Mar 12th 2007 6:40PM]

I think that the future of motion capture is to get away from doing it visually. Imagine the kind of data that you could get by strapping Wiimote like accelerometer devices to a person. And by getting away from needing a special screen or lights and cameras you could actually do mo-cap in real world locations that are similar to the in-game or in-movie locations. A mo-cap actor could actually deliver his performance just like a regular actor and at the same time as the regular actors. Furthermore if the devices are waterproofed they could be used to capture motion underwater which is probably extremely difficult or even impossible to do with current techniques.

[Moby Disk @ Mar 12th 2007 8:35PM]

This company was the only one at GDC who didn't have a live demo. I think that it requires too much post-processing and manual cleanup for it to be possible to do in real-time. It's really cool to know that machine vision has gotten to this point, but for the foreseeable future, but I doubt this will ever be able to compete with using markers.

[ethana2 @ Mar 13th 2007 2:03AM]

How long till it goes open source? We better get working on that...

Multiple cameras with some way to register position and orientation...
Fully documented light sources for easy engineering...
Recognition of specific features and subsequent vertex mapping...
Joint mapping...
Soft body form recognition and mapping...
Texture mapping...

Sounds like a fun project. Maybe all we need is a blender plug in...
I've always wanted an accurate, virtual self. But I'm weird like that...

[Zut @ Mar 13th 2007 7:56AM]

This isn't really groundbreaking! A friend of mine did this as a research project at Uni!

[John Venzon @ Mar 13th 2007 3:50PM]

If you listen closely, off in the distance, you can hear Andy Serkis screaming.

[shaun @ May 26th 2007 1:11PM]

Ok another company to check out is www.phasespace.com. They use active led lights as markers, the nice thing about active leds is that you dont need a suit just attach and go. also since they make their light falhs in a particular patter the cameras and computers can generate realtime images of objects movin faster then the human eye can percieve. check them out, its som pretty cool stuff.