Boffins fill a dome with 480 cameras for 3D motion capture Move over Gollum: Carnegie-Mellon boffins are working on 3D motion capture that works without tracking sensors, and can pull together images from hundreds of sensors. Their work, documented in full here, goes way beyond the motion capture tricks that let Peter Jackson give Gollum and King Kong to the world: they put their …
For example, hundreds of camera-phones capturing live sports could become the input images to MAP, rather than the controlled environment of the Panoptic Studio.
Will MAP automatically ignore all the phones being held in portrait mode? That should cut down the processing load to barely anything at all...
you do have gps, exif, wifi and bluetooth data, being recorded/transmitted on smart phones.
I don't see it as being easy, but possible, imagine I enter a stadium via gate 'f' smart phone in pocket, the time and location could now be known, but even if I was pointing my smart phone in the right direction, it is improbable it will capture anything worthwhile, as reference to my location, via mapping of the stadium shows this. However those that enter through gate 'c' should have a good view, with gate 'a' also possible.
You could also grab facebook and twitter for the event/incident and match up the data.
post mortem, you will certainly have something worthwhile, whether this is a 3d image or not, I can see law enforcement, marketing, sale etc. loving the idea, time to get my tinfoil hat :-)
Try looking at the point-cloud for this
http://photosynth.net/view.aspx?cid=0ef05319-4b7b-491a-8e75-040c1af4ce56
which I believe is generated purely from the photographs.
More photographs = more accurate positioning.
This studio seems like an application of the same sort of thing for movies
But could you capture confetti, and track multiple objects.
Not saying it doesn't sound extreme, but back in 1999 i worked on Ir motion capture and even then you got odd frame drops in the middle. this sounds more like a always working no drops perfection type of deal.
I wish i had stayed in the mocap industry.
A couple of comments about 480 cameras.
First, making a probably-inappropriate analogy to GPS satellites, more is better. You can figure out your altitude, longitude, and latitude from 4 satellites overhead. You firm up your position considerably by being able to spot additional satellites. Further, more than 4 satellites helps compensate for interruptions by trees, terrain, buildings, weather, etc. The alternate GPS constellations help further.
Similarly, additional cameras help (at this time) because - per the article - help see past blockages. A few cameras are sufficient for now when you've got a single actor Golluming around the sound stage looking for his precious, but putting a dozen actors and complicated terrain in the mix will block line of sight from multiple cameras. (Not that a dozen actors will fit in this trial unit.) More cameras are more chances to bypass blockages
Second, the application benefits from 480 cameras *now*. The lessons here are probably going to go in two directions: harnessing more cameras (per the numerous suggestions in comments to integrate lots of public cameras), and streamlining a studio rig to fewer cameras. Technology seems to move that way frequently: the first time a new capability debuts, it is overengineered and complicated. Successive generations of the product get simpler as engineers build experience and figure out to milk more from less.
The Register 
Biting the hand that feeds IT
Copyright. All rights reserved © 1998–2024
