I've been very busy the last two weeks, sorry for everyone I've yet to reply to. Basically, I'm in the middle of about a dozen different projects. Working on a dozen simultaneous projects is a little more than my usual half-dozen, and since no one really pays me well enough to just concentrate on one big project, adding to that the fact that most of the current projects are pro bono, well... everyone will just have to wait in line. sorry ;-)
The great news from all this activity, is that Stéphane showed up for our latest session Sunday afternoon with a working version of the Processing OpenCV library running on Windows. This was just in time for the two-week long rehearsals we started today, in order to finalize Wolf Ka’s latest dance piece. Since Wolf wanted to run everything off of a PC, it was a good excuse to give our OpenCV library a second real-world test run before unleashing it onto the wild. Especially since we have to have this piece ready by early April.
We found a few crazy things just in time, and will probably find a few more, but so far everything is running great, almost immediately, even better than I expected, and we were able by the beginning of the first afternoon to start working on the dance+algorithms, as opposed to plugging things in and banging on the machines. I'm thrilled as to how quickly you can get a full kick-ass tracking program up and working with an absolute minimum of code. Totally cool, that. It's all in line with the attitude of offloading the ugliness -- Processing-style -- and just getting to work on the essentials; but not too much, making sure not to dumb things down.
We've also taken the opportunity of going through all the original OpenCV examples and documentation, compiling directly in C/C++ through GCC, just to understand how the library works at the most basic level. As it turns out, it's not all that hard, but we've just started a new phase of this research, so we’re still only just getting started -- which is pretty exciting for me because I’ve already begun to imagine million little crazy things I can do with this library. Anyway, things like face tracking are an absolute shoo-in for the first beta release (thanks Guillaume), as well as movie import (.avi), and some simple brightness/histogram magic for treating images -- either before you analyze them, or just because you want a fast Photoshop style filter inside of Processing. We're also working on object tracking, but I haven't gotten around to Kalman predictive filters, as was suggested here in the comments, but I'll get around to it -- it's just a question of figuring out how the various available examples approach the problem and making it a natural integration into Processing. A lot of Stéphane's and my work have involved debating how to make the whole process very Processing-like, both simple & modular.
Oh, I should mention that the rehearsals with Wolf are taking place at the beautiful Pavillon Noir, the national dance center recently created by/for Angelin Preljocaj. If you've visited Aix-en-Provence recently, you should know what I'm talking about, it's absolutely unavoidable. Since this time we are working with dancers, it nice to be in an actual dance studio, as opposed to some dingy basement at the Art School.
I should also mention that I have two great students with me, Stefan Schwabe who is currently on loan this year from the design school in Halle, and Fabien Artal, a veritable personnage of the Atelier over the last three years. I’ve been lucky these past few sessions to have a steady stream (but small enough to be manageable) of extremely competent students who transfered to the school with a very specific understanding of what it is we’re trying to accomplish, which is very important in my case, since the Atelier Hypermédia is not always the easiest thing to explain to an art student. It is actually thanks to this current push of students that we have been able to take on some of the more ambitious projects of the Atelier.