https://github.com/gnebehay/OpenTLD
What have I been doing?

Yeah... that's how it feels sometimes..
I downloaded the binaries to see what it would give me..
Unzipping it on a windows box .. and what do I see ..
Ahhaaa ! Some very familiar OpenCV libraries...

core - for core opencv stuff
highgui - for the screen
imgproc - for image processing
video - for the camera
tbb - i think is for python .. dunno why they'd need that
Qt for portable windowing/display system usually.

And here we goooo !

This exe gives you a bunch of options to fill out first..

The sad thing is .. if the mouse was supposed to work on selecting a region in the video .. it doesn't work.. at least for my computer.

I manged to eventually create an initial bounding box with my eye in it.

Results = Pretty Good 

About on the same level as Face Detect in OpenCV with FIND LARGEST FACE setting on.
However, there some important differences here.  The biggest one is that this is an arbitrary region selected, with no previous training data .. Which is really the whole point of TLD - that the (L)earning is being done and accumulated each frame.

This would be very helpfu in the
"What is this?" game :
You have a camera running an OpenCV background subtraction filter.  You hold up something new in the camera field of view, it switches from background subtraction to TLD+ .. MRL would then ask "What is this" .. You would say, "A rabbit"..  MRL would say, "Ok"....
Next time you hold the rabbit up, MRL would say "I see a rabbit"

Google Goggles or Google image search could be used too - but we'll add that later.

Ok - so we definately want it in MRL... how to do it?

Possibilities on the Fastest way to get into MRL.
1. Recompile it with OpenCV 2.4.6.1 - and use JNI or JNA to interface with it. 
Pros - Quick to implment Cons - Hand tool C interfaces to C++ code, have to rebuild 8 different platforms
2. Use JavaCPP
Pros - Quick & Clean  Cons - have to rebuild 8 different platforms
3. Re-implement in Java
Pros - no native code !  Cons - longer to implement

Well, time to get started with step 0 - "Use the Source Luke"

This should keep me busy until I get a working kinect :D