Forum

Contour vs. TBeta vs. Reactivision vs. Snowflake

Hi all,

i’m working on a multitouch project and i am wondering wether there are any reasons to use additional tracking software instead of the Contour plugin. i know, the AsVideo-node is some kind of bottleneck, so i try to avoid it by doing image processing stuff like color filtering and blurring with ffdshow in advance and all the camera calibration and undistortion stuff AFTER Contour. this approach works well in terms of performance and latency. calculating some additional params like speed and orientation of the finger touches/movement is fortunately not that difficult with vvvv. so are there still any advantages using eg. TBeta or Reactivision over Contour in v4? please tell me your experiences or maybe any issues i didn’t take into account…

thx!

flux

Contour is a bit of a cpu hog… Dedicated trackers are a bit faster, but you have to put up with running different programs…
Swings… Roundabouts…
:)

advantage of doing the tracking in vvvv is that you can vary the filters in the tracking chain. (eg chaining contour, trautner, fiducials, facetracking etc)

disadvantage is that each freeframe/directshow plugin introduces quite some latency even if it does nothing. (tested that with a dummy filter) + as catweasel said it uses quite some cpu, where the latest release of ccv is really fast.

also note that theres currently no filternode in vvvv which removes the blury parts of an image which is needed for di - tables.

i think the best way to do any video analysis nowadays is to write a dedicated openframeworks app and hand over the result data by osc.

that way you can adjust the captured stream to your needs (b/w, gamma, sharpen, threshold, etc) and also use all the tracking you want on the same stream (contour, blob, color).

also it gives you the freedom to use threading if you run into cpu issues or even run the tracking on a dedicated computer.

also, it is fast as hell.

hi velcrome, but for openframeworks you need to code ;-)

about this subject, i experienced in jully 2009 bad very bad troubles of “out going pointers” with Tbeta and the TUIO decoder dll, wich is third party.

It was more flexible and strong to do tracking with vvvv than in fact adding other softwares…

There are a lot of good things about the video capabilities of vvvv, and I certainly don’t want to disapprove of them. Subframe precision is something that most software tools don’t offer. But there are tendencies of unstable behaviour and also strange lag with filters.
Of course you can dive into it with tools to debug the filter graph, but somehow it seems easier with OF lately. Yes, you need to code, but they did a good job to keep away strange pointers and other c-voodoo from users.

Don’t get me wrong. I still hate compiling, still don’t like semicolons and even less strange syntax and header files. But for video analysis OF is a real option for me, as it gives me more freedom with openvc.

alalala, if OF was more stable at the time i began this project:

http://www.le-chat-noir-numerique.fr/

surely i would have written it with OF… but now its too late, too many things written …