Hey, I am conducting an experiment for my master thesis using the pupil core headset. I want to investigate the driver's distraction in a multi media system. First one using touch versus a a "normal" system setup with a central "joystick". For further detail: The actual Mercedes CLA vs. C-class. I'm an absolute beginner in this field and would like to know what you think is the best calibration setup in this case. I would suggest the calibration has to be at that distance the screen of the multi-media system is located, right? But which method is the most accurate? I am using the newest software 3.0.7. When conducting the normal screen calibration, holding my tablet right in front of the car-display, I constantly get 15-25% of pupil data missing due to confidence <0.8. This seems pretty bad for me. I know the data confidence changes for different type of eyes and depends on size of eye lashes etc.. But my eyes are pretty much normal I would say, so do you see any possible improvement here or do I have to accept that error? Of course this does not affect my thesis degree, as the department doesn't know it any better but it's kind of disappointing to have results this bad. Thanks for any kind of help :).
You likely saw this already - but wanted to make sure it is linked here again: https://docs.pupil-labs.com/core/best-practices/#best-practices - in many cases low confidence pupil data is due to the physical adjustment of the eye cameras. When you say data is dismissed due to low confidence; you are talking about during calibration?