(email written to Southampton, 4/11/2014)

Requires Boost ASIO

Following Andreas' suggestion, I managed to get Boost ASIO working. It 
looks like the best principle is that the process that generates data 
(in our case, the head tracker) works as a server, and any user of data 
works as a client.

So I've wrapped one of my simplest trackers as a UDP server. It now 
transmits bounding box coordinates as a string.
The client simply reads it and displays in on the screen.

I put a very basic client-server demo at
http://cvssp.org/svn/s3a/Stream3/Code/Network

My first face detector demo (which was wrapped as a server) is at
http://cvssp.org/svn/s3a/Stream3/Code/Network/ViolaJones

Both run in Linux. They should compile on Windows as well, but I haven't 
tested that. Please let me know when you manage to compile the basic 
client program under the first directory (above).

Once compiled, all you need to do is run the server from the command line:
    server
and then the client with this syntax
    client <host name> <port number>

In the server code, I'm using port 8888, so if you run both programs in 
the same machine, the syntax for the client becomes
    client localhost 8888


You can also try to compile the face detector, but that requires more 
libraries, so it might be trickier. But in principle you shouldn't need 
to worry about that. If I compile it statically here, I can just send 
you the binary file and it should work fine for you. This program just 
requires a webcam. We're still working on the Kinect2 version, but the 
above should be enough to sort out the integration.

At the moment, the face position is streamed in this format:
<object label> [<size in rows> x <size in columns> from (<top row 
index>, <left column index>)]

For example, if a frontal face is detected, you get this line:
fontal_face_0    [160 x 160 from (270, 206)]