Introduction to Computer Vision

The intention was to adapt my TPO Farfalle patch for an IR camera and projector in Isadora.

The first order of business was to download Andres Colubri’s Syphon Syphon Library in Processing 3.

I used circle tracking in Isadora to receive the feed from the IR camera. This meant feeding that the Syphon receiver actor (for which Processing 3 has to remain open and playing for it to work) into the Zoomer actor. Since the IR camera could see more than the projector’s playing space, the Zoomer actor had to shrink the image down to fit within the boundaries. This was particularly tricky due to the fact that tracking a person is dependent upon their height. The wall made setting the place particularly difficult. The Zoomer actor was then fed into Horizontal Flip, Difference and Contrast Adjust actors, which was then fed into a Video Mixer actor, along with a second input from the Horizontal Flip actor, and then finally into the Eyes actor.

From the Eyes actor I used the horizontal and vertical positions, as well as the object velocity, all of which went through a Smoother actor. The GLSL Shader used the horizontal and vertical positions and the velocity defined the saturation level, as it did in the original patch. Screen shots of the patch can be seen below.

Here is a link to a video of what the patch looks like: Note that it is slight ahead of my feet due to me bending over. It is tracking the top of my head.

I also attempted to incorporate two objects into the patch using Eyes++. I did this by adding two Blob actors and using the Calculator and Limit-Scale Value to average the position between the two objects, as well as average their velocity. While I was able to get the correct readings and it worked in theory, the patch does not lend itself to using two objects and I did not like the practical result. Screen shots of this patch cam be seen below.



Post Class Update:

Rather than using the Circle Tracking technique, I have updated the Isadora patch to use Background Subtraction. This has allowed for a more stable tracking of the object/person in both the single and double version, taking into consideration their extra movements, and has made the double object version able to work when one or both of the objects/people are stationary.

Here is a screen shot of the Background Subtraction (single object version):

Here is a screen shot of the Background Subtraction technique in the entire patch (double object version):

Here is the Isadora Patch:

SBM/TPO Background Subtraction.izz

Here is a video demonstration: