InfoCentral opened this issue on Mar 16, 2006 · 20 posts
InfoCentral posted Thu, 16 March 2006 at 8:05 AM
Attached Link: http://www.naturalpoint.com/optitrack/index.html
Has anyone seen this advertisment? It certainly is cheep enough but I believe you need to program a front end for it to work with MotionBuilder or Life Forms. Hummmm...looks interesting!
Miss Nancy posted Thu, 16 March 2006 at 1:04 PM
it's ingenious! my experience with IR-videocams indicates that 8 - 12 feet is the optimal range for those LEDs, which might be good for an entire human body, depending on focal length of the lens. I'm assuming they're using cameras with moderate light sensitivity (e.g. 0.5 lux) so that they get good contrast between the balls and the scene (greyscale).
archdruid posted Thu, 16 March 2006 at 1:17 PM
From what I saw, it looks like you get an SDK, one camera, and a set of IR reflective spots. there doesn't seem to be any program FOR it. The thing that catches my attention most, is that, since it's optical, you're GOING to need at least three cameras. By my calculations, by the time you're done with developing the 'ware, and getting more cameras, you might as well go with something more developed.... something with it's own software. Lou.
"..... and that was when things got interestiing."
InfoCentral posted Fri, 17 March 2006 at 7:56 PM
Even if you buy an additional 3 cameras ($900) it still would be very cheep. Investigating further I find that you will indeed have to write the program to record and use the data this produces. This is a major problem or could be a major opportunity depending on how you look at things. If you were a programmer and could write a front end to get the data into Life Forms or Motionbuilder you would have produced a much needed program and can cash $$$ in on it. For somebody like myself this product is useless, even thought very cheep, until someone does this and markets it. Im sure the market would be huge for a low cost personal motion capture device. Hummmmmm...I think I'm going to pick up that QBacsic book again and start learning...this could be that million dollar product I was looking for...now where did I put that book...
yoshi-mocap posted Sat, 03 June 2006 at 8:26 AM
Hello,
I am developing an exactly such application.
http://www.geocities.com/mocap_is_fun/
It can capture a single person full body motion and output to bvh format. It is still in development, tracking is failing when motion is too quick.
nemirc posted Sat, 03 June 2006 at 11:26 PM
any chance you can show a "behind the scenes" kind of thing?
nemirc
Renderosity Magazine Staff Writer
https://renderositymagazine.com/users/nemirc
https://about.me/aris3d/
dueyftw posted Sun, 04 June 2006 at 12:04 PM
Thoes cameras have no software support. . Syntheyes has mocap support and is only 400 dollars.
Yoshi; I would love to help you with your programming but I have no "Class"
Have you tride strobe lights, filtered with a lens to match the collor of the markers? That might remove most of the false readings and alow faster tracking.
Dale
yoshi-mocap posted Sun, 04 June 2006 at 1:21 PM
Quote - any chance you can show a "behind the scenes" kind of thing?
When I get a chance, I'll put the images of camera rigs, reflective markers, velcro attachment, etc. in my site.
yoshi-mocap posted Sun, 04 June 2006 at 1:28 PM
Quote - Thoes cameras have no software support. . Syntheyes has mocap support and is only 400 dollars.
Yoshi; I would love to help you with your programming but I have no "Class"
Have you tride strobe lights, filtered with a lens to match the collor of the markers? That might remove most of the false readings and alow faster tracking.Dale
Yeah, I meant the tracking fails when motion is too quick and markers travel further distance than my algorithm expects, not so much about the OptiTrack camera failing to detect the markers.
dueyftw posted Mon, 05 June 2006 at 12:59 PM
Oh, Have you tride to rewrite algorithm in assembler code to speed it up?
C++ can be slow, real slow if you are relying on functions to perform tasks in your algorithm.
Dale
Miss Nancy posted Mon, 05 June 2006 at 1:26 PM
apparently one needs 6 cams to take care of marker occlusion. see yoshi's thread in poser forum.
dueyftw posted Wed, 07 June 2006 at 12:20 AM
I have been wondering why are you having problems with fast motion tracking? Is your program generating BVH as it tracks? If so, maybe a two pass approach, with one recording, the second rendering the BVH? This way your not limited by the computer or program speed.
Dale
yoshi-mocap posted Thu, 08 June 2006 at 1:54 AM
No it's nothing to do with C++ language or speed of program. In order to track markers from one frame to next, there has to be similarity between them. But if motion is too quick and markers move a lot, there will be less similarity.
dueyftw posted Thu, 08 June 2006 at 12:44 PM
Sorry, just trying to help. Syntheyes has the same problem with very fast tracking, so the program allows you to track points by hand.
Dale
Miss Nancy posted Thu, 08 June 2006 at 1:36 PM
"similarity" may refer to the problem that the algorithm confuses nearby markers with each other, if the distance traveled by a marker between frames is >= to the distance between nearby markers. hence dale's technique. it might be tedious at 120 frames per second (or faster) and a minute of motion per file, but at least an human can tell which marker is which with less difficulty than a computer, if all the markers are identical in size, shape, labelling, colour, etc. one might need "smart markers" to more easily automate that, or else make sure the frame rate was fast enuff to keep marker movement small, compared to marker spacing.
yoshi-mocap posted Thu, 08 June 2006 at 11:39 PM
Quote - Sorry, just trying to help. Syntheyes has the same problem with very fast tracking, so the program allows you to track points by hand.
Dale
Have you actually tried to use Syntheyes to capture human motions? It looks difficult according to some of the posts in their forum. > Quote - hence dale's technique. it might be tedious at 120 frames per second (or faster)
and a minute of motion per file, but at least an human can tell which marker is which
with less difficulty than a computer, if all the markers are identical in size, shape,
labelling, colour, etc. one might need "smart markers" to more easily automate that,
or else make sure the frame rate was fast enuff to keep marker movement small,
compared to marker spacing.
It is already possible to manually identify the markers. But as you say it can be tedicous soon if you have to do it often.
InfoCentral posted Sat, 10 June 2006 at 10:41 AM
I went to the website but I didn't see anywhere where it was stated that Syntheyes was capable of producing character motion capture. Camara tracking yes, but character motion capture no.
dueyftw posted Sun, 11 June 2006 at 1:21 AM
http://www.ssontech.com/synsumm.htm
Motion Capture – for faces or whole bodies, using multiple cameras.
How good is it, I don't know. I have used only the camera tracking demo. But I don't see a bvh export.
Dale
InfoCentral posted Sun, 11 June 2006 at 9:20 AM
Quote - But I don't see a bvh export.
Dale
Yeah, I didn't see it either. Without that I don't see how your going to attach the motion to a character?
Fulvio posted Wed, 19 July 2006 at 6:51 AM