Just wondering why you rolled your own gesture recognition system rather than using something off-the-shelf like the Leap Motion.

Also, your comments about the Python environment are maybe a little unfair... You have an unusual (and some might argue overly-complex) use case where you're combining it not just with the third-party OpenCV package and wrapper, but also Tornado and Node.js. It would have probably been easier going with either the Python approach or the JavaScript approach rather than combining them.

All in all, pretty neat though.