Kinect + QC

Marshall B's picture

Hey everybody, I'm relatively new to QC and I was wondering if anybody had any recommendations on how to utilize the Kinect in QC. I'm running this off my macbook 10.6. Thanks!

jstrecker's picture
Re: Kinect + QC

Anything specific you want to accomplish?

The tryplex OSCeleton plugin is fun.

jstrecker's picture
Re: Kinect + QC

Also KinectTools.

Marshall B's picture
Re: Kinect + QC

Thanks very much for this. I am interested in using the kinect to cause separate animations to move. For instance, If a person walks into a space, their image will effect other animations (simple 2D mov files). I plan on running multiple files through modul8.

gtoledo3's picture
Re: Kinect + QC

Here's a long winded bunch of info :-)

I'd recommend kinect tools for scenarios where you wish to use more than one kinect; it's the only plugin for QC that I know of that successfully does this. However, Kinect Tools will have an issue when it's connected to a Consumer patch in QC, and you turn that consumer of/on. The kinect won't want to "fire off" again... I usually keep a Billboard constantly evaluating in the background, connected to the depth image, to workaround that quirk.

I'd suggest investigating v002's kinect plugin for scenarios where you wish to obtain the infrared image. The depth image also has a greater bit depth than kinect tools (there are more steps of resolution in the "z axis" than what is delivered by kinect tools...maybe my semantics are wrong here, since I don't know the kinect tools code, but it can be visibly seen).

This plugin, which you will need to compile in Xcode, is also available: https://github.com/stoulouse/Quartz-Composer-Open-Kinect-Plugin

...is really promising, and has a cool feature that delivers the "point" that's nearest to the camera. That can be really cool for scenarios where a user is just going to be waving their hand in front of themselves; it's like having a simple blob tracking built in. That said, it doesn't address multiple kinects simultaneously (in my experience), and it has some issues with not "letting go" of the kinect once it's been started up, but is "ok" if you're just running the composition. I'll try to make a "fork" of this on github sometime soon, because I cleared up some of the issues with this plugin awhile ago... and haven't thought about it until now!

I'd also suggest looking at installing : http://code.google.com/p/tuiokinect/ - A great, simple, blob tracking app for the kinect, and the current tuio plugin for QC : http://www.tuio.org/?software

That will deliver a "structure" a information that you can then use to place objects wherever the "blobs" in a scene are... somewhat similar to the OSCeleton kinect stuff, but not a skeleton, and with no calibration time.

For rendering the depth image as a mesh or point cloud, you might want to look into GLSL or OpenCL examples in the composition repository for the kinect, or for "height fields", as well as Kineme3D's height field patch.

Marshall B's picture
Re: Kinect + QC

wow! Thanks for all of this great information. I'm still new to this so it may take me a while to figure something good out. I'll document what I come up with and post it asap.

Marshall B's picture
Re: Kinect + QC

Hey again, Is there anybody who could help me with setting up a Kinect to affect video speed? For example, I made a movie of a tunnel in After Effects and I would like to set up the kinect to allow the viewer to "swim" through the tunnel. Any advice??

gab_lecup's picture
Re: Kinect + QC

hello everyone. i’m quite new to QC and i’m going crazy with all its possibilities, but i’ve been hitting a hard brick wall on this one subject, so i finally decided to ask the wise: i just can’t seem to find a way to get a point cloud from the kinect depth view. For now i get the depth view through a Syphon feed from Delicode Ni Mate, and send it through a 3D HeightField and a 3D Object Renderer all in a Polygon Mode. It works fine but i can’t get rid of the extrude effect (i can’t filter out the depth to get just the points of the body). i’ve tried the Kinect Primitive Points by _1024 and it renders the points, but i can’t extract the xyz structure to render it as gl points … i’ve seen many example videos lying around this forum (lots from gtoledo, wherever i look you appear) so i know it’s possible. From what i’ve gathered there should be way to do this using GLSL shaders or OpenCL, but i haven’t found anything specific … I also think i understood the depth info is coded in the alpha channel from the depth view, but i’m thinking there must be a better than transforming depth values to an image and then back to depth values, can’t i just access the point cloud? like xyz coordinates for each point? i got this to work in a click in max/msp/jitter. i don’t even need the color values … well there you are, i’ll be incredibly thankfull to anyone who can put me out of this misery … and if i really have to learn about shaders to get this going, please tell me where to start, as it seems a bit steep on my own. thanks loads to all of you, i’ve been spending all my time on this forum and you all rock. cheers,