For my final project I want to use rough posture recognition to create system that triggers a photograph to be taken as soon as a pose enters a certain set of parameters. In the above photographs I attempted a rough approximation of this system.  The photographs on the top are photographs taken when faceOSC detected that eyebrow position and mouth width equaled two, the second row is when the system detected both those parameters equaled zero, and finally the bottom row shows a few samples of the mess ups.

I have a couple ideas of how I might implement this project at varying levels of complexity:

• trigger a DSLR camera whenever face or body is in a particular position. Make a large format print of faces in a grid in their various positions. 
• Record rough face tracking data of a face making a certain gesture. Capture that gesture frame by frame, and then capture photographs that imitate that gesture frame by frame.
• Trigger photographs to be taken when people reach certain pitch volume combinations. Create an interactive installation that you sing to and it brings up people’s faces that were singing the closest pitch volume combination.

All of these ideas involve figuring out how to trigger a DSLR photograph from the computer and storing a database of images based on their various properties. Here are some resources I have come up with to help me figure out how to trigger a DSLR:

• using arduino (tutorial page)
• Using an open frameworks library
• Communicating with an application that already talks to the dslr. PC only ):

In terms of databasing photographs based on their various properties, Golan recommended looking into principal component analysis, which allows you to reduce many axis of similarity into a manageable amount. He drew me a beautiful picture of how it works:

I also found Open Frameworks thread that pretty much described this project. Here are some of the influences I pulled out of that:

Stop Motion by Ole Kristensen

Cheese by Christian Moeller

Ocean_v1. by Wolf Nkole Helzle