Skip to content

06.13.2009 – Motion Tracking

It has been so cool working on motion tracking, but it has not been without its problems. The code has actually been fairly easy and straightforward, but I’ve hit the ceiling of what my RAM can handle. And I’m on a 2GHz MacBook. By the time I’ve captured the video and processed it to recognize and assign tracking nodes, I’ve used up 50% of my RAM. I didn’t realize how big a problem this was until I started trying to code in gesture recognition. For every functional algorithem, I lost surprisingly large amounts of processing power. I’m not sure if it’s because 25 times every second I’m performing such a huge number of mathematical computations or the camera driver just sucks up all the power on its own. And it’s not just a problem in Pd: I’ve tried using a tracking library in Processing, but it actually seems to consume more power than Pd. What I need is a super efficient algorithem that will give me enought headroom to do all the cool gesture recognition stuff. I’m looking into trying yet another program that is from the offset more geared towards video, but I’m not completely convinced it will be the answer to all my memory problems. My final solution may actually be to string together three different computers: one to capture input, one to perform the evaluations, and one to run the audio. Maybe four so I can do more with graphics animation…

Anyone have a computer they would like to donate to the Aaron Fryklund Center for New Music Research?