Home › Forums › Software Development › NaN data from GazeDataRecieved
- This topic has 8 replies, 3 voices, and was last updated 3 years, 3 months ago by Grant [Tobii].
- 08/01/2020 at 21:36 #13804
I am receiving gaze data that, for a series of sequential GazeData points, all values are NaN.
I would like to know what that implies about those data points, as compared to times when i am receiving no data from the tobii at all.
It seems like it would make sense that the tobii is not reporting any gaze data(even empty) when if cannot ‘see’ any eyes to track, and all NaN entries would imply it ‘sees’ some sort of eyes but is unable to determine gaze data for them.
What i’m driving at is: what’s the difference between receiving no GazeData(not even an empty object) when you would expect a datapoint to come in via GazeDataReceived, and receiving the GazeData but all the values are NaN.
A caveat is i’m adapting from legacy software with low confidence. If tobii is initialized correctly, should it always broadcast GazeData even if it cannot register a viewer? The old software would report nothing if there was no person in front of the tracker, and i’m unsure if that is how the tobii operates or something in the legacy app that is swallowing the ’empty’ data points.
I’m using a 4C Tracker with the Pro SDK in .Net
– if i’m in the wrong place please for the love of god tell me where I supposed to ask this because I’ve tried twice now and i’m perpetually in the wrong place.08/01/2020 at 22:52 #13806
inb4: I posted here a bit ago and was told to go to pro support, then i posted in the pro support portal and they told me to come here. I have no idea where i’m allowed to ask questions.09/01/2020 at 17:53 #13810
Hi @scaustin34 and sorry to hear about your troubles and the run around you have received. The issue is that the Tobii 4C is part of the Tobii Tech business department whilst the Tobii Pro SDK is part of the Tobii Pro business department, where we each have our own unique support structure.
Accordingly, my suggestion firstly is for you to try (at least experimentally) running our own SDK which are designed for the 4C Range such as our new Interaction Library which has .NET bindings and you can download @ https://developer.tobii.com/consumer-eye-trackers-interaction-library/
It comes with a sample ready to run directly (We recommend VS2019 for this) and should allow you to quickly see if the NaN’s are appearing in the gaze stream here as well. If they *are*, then that may imply an issue with the hardware itself as Nan is returned when no eyes are found or there is some issues of connectivity.
Let’s go through this stage first and report back and see where we are. May I ask if there is a particular reason for using the Pro SDK .NET Bindings? Certainly, if you are using the Tracker for analysis or long term data storage, you will still need the analytical use licence, which I assume you have, but you can still use another SDK if suitable. And as the Interaction Library support C#, you should hopefully not face much issue in code transport.
Thanks for your patience whilst we try to get to the bottom of the issue.10/01/2020 at 16:24 #13815
I got the interaction lib sample stream project going and if i’m looking off screen i’m receiving 0 values if i’m looking off screen/blinking, but not NaNs. I guess that should indicate there’s a problem with our application that’s causing the NaNs?10/01/2020 at 17:46 #13823
I had this whole reply typed up and it isn’t posting. Not sure what to do about that.10/01/2020 at 17:47 #13824
I guess i’ll try and post it in parts because smaller messages seem to post fine. Were using the tobii research x86 nuget pkg now which to my understanding is part of the pro sdk. The app is running its own calibration and then listening for the gaze data received event. Would the 4c sdk be capable of that?12/01/2020 at 20:11 #13830
Hi @scaustin34, okay thanks for the feedback. Good to know that the interaction library API is working as it should as we can rule out that you have a hardware issue which is also a positive development.
Calibration with the 4C is generally required to be ran using the Tobii Core Software System Tray Icon, so perhaps the calibration routine is faulty? Did you try using the tray icon then go straight to your Pro SDK program?
It would seem in any event that the issues does indeed lie with your project code, which unfortunately is not something we can offer detailed support on. My suggestion would be to run one of the Pro SDK sample applications to see if the issues persists there… assuming it does not, then further examination of your code would be required.23/02/2020 at 17:19 #13969Hongbin ZouParticipant
There can be Nan data from FixationDataStream even when tracker is working properly.26/02/2020 at 11:05 #13973
Thanks for your input Hongbin.. this is indeed true in the situation where the Eye Tracker is working correctly but the user may be looking outside the screen boundaries but in general if NaN are received when the user is looking directly within the tracking boundaries, then this may be cause for concern.
- You must be logged in to reply to this topic.