The computer I am talking about above is a Surface Book running Windows 10 Build 14393. It was actually not running “perfectly fine” a few times. When launching sample 103 app, sometimes it would take about 5-10 seconds before any valid head tracking data was shown. Then after about 10 executions, I began seeing error (TOBII_INTERNAL_ERROR code 1). I have tried unplugging and reconnecting the Tobii 4C tracker. I have tried power-cycling my Surface book. Still same error.
I took my 4C and connected it to a Lenovo Thinkpad running a Windows 10 Insider preview build and Tobii’s Core eye tracking hot fix release 18.104.22.16858. Details found here: https://blogs.windows.com/windowsexperience/2017/08/02/announcing-windows-10-insider-preview-build-16257-pc-build-15237-mobile/#45L4HP9Ygw1mZ38l.97
I built the https://tobii.github.io/stream_engine/#example-26 example using VS 2015 as a Win32 Console program, and ran the program on my Lenovo Thinkpad and I am able to see head-tracking data (after about 5 seconds of launching the app). The second time launching the app, I get a code 4 (TOBII_ERROR_NOT_AVAILABLE) – I suspect this is because when I closed the application previously the stream was still open? I run it again, now I see no error and the head-tracking data is displayed on the console immediately.
Head tracking seems a little experimental still. Could someone please provide the algorithms to take the eye positions, and convert it into the head position and rotation data? I’ve not had any problems with the eye gaze, eye positions, or eye fixations and it seems more reliable. Though it doesn’t seem hard, but the fact that Tobii created a separate stream for head-tracking indicates there’s something special that an algorithm which merely uses eye positions would miss out on.