Hi
I have some experience with the EyeTribe(ET) and I wonder whether anybody has similar problems as I do/did, so, my questions are:
How many times do you repeat the calibration until you have a reliable one?
How long are your sessions if you use it in user studies? or
How long can you collect data with the ET? Does it automatically shut down after a while?
What is the viewing distance?
Do you use any extra tool to stabilize user's head or to preserve the calibration and the viewing distance?
Do you use ET more for post analysis or for real-time interaction?
Thanks
Hi Kenan,
I used The Eye Tribe-Tracker for three experiments with a total of about 120 participants. The combination with ogama 5.0 software for post analysis, which I used, worked well.
The mean session time for using this combination was about 20 minutes.
I used a 12 point calibration with no extra tool to stabilize user's head and had to exclude about 8% of my participants because of low data quality (e.g. mascara-problems..)
The viewing distance was 60 cm using a 1280x1024 Display.
Repeating of calibration:
In about 70% i had to calibrate only once..
in about 25% i needed a second one.
in about 5% i needed more than 2 calibrations
I had some shut down problems with my EyeTribe-Ogama combination, when the mouse curser left partcipants screen.
May be shutdowns depend on computing power. I have an i7-machine and no further problems even with long session duration.
Best regards,
Matthias
Have used the EyeTribe to check whether participants maintain fixation during the parts of an experiment where this is crucial. Experiment lasted over 1 hour, and usually a single calibration was enough to get the thing going. I did implement an automatic re-calibration when drift got out of hand (> 2 degrees). Viewing distance was around 5 centimetres, and a chin rest was used (why would you not?).
How long you can collect data depends entirely on how you program your task. I used my own Python implementation (see attached link), which works fine. If the EyeTribe shuts down before you want it to, chances are you failed to keep sending it heartbeats and/or commands (if the EyeTribe does not get input from your PC every so many seconds, it will break the connection).
As you might have gathered, we used the EyeTribe for both online gaze interaction (to start a calibration when needed), as well as offline analysis (to check the location of fixations).
http://www.pygaze.org/2015/06/eyetribe/
I have used EyeTribe in two different setups:
SETUP 1:
A pilot study with 6 users and a main user study with 20 users.
Task: Gaze based interaction with images.
Equipment: Desktop PC, EyeTribe, chin-rest support and game controllers for maintaining additional interaction.
Explanation: The experimental design had 3 to 5 blocks of comparable trials.
The blocks took about 3 to 12 minutes each and they were separated with mini breaks (1-2 minutes). Average duration of the experiment was about 80 minutes. The complexity of the tasks(trials) was relatively low but the experiment was rather long. Therefore we wanted to put some breaks between each block which in return made us re-calibrate the participants.
Challenges:
1) During pilot tests we noticed that the aluminum case of the eye tracker was getting hot. In random intervals it was also falling into a deadlock (multi-threading).
Then we had no choice but restarting the eye tracker. The well-designed orchestration part allowed us to continue from the trial where we had restart.
But some participants have experienced many restart/recalibration.
We are still not sure whether that is related to the SDK, hardware or our implementation.
2) The time for training some participants were longer than the others. Depending on the target group this might be a bad sign or a useful indicator. I also designed similar studies with more sophisticated eye trackers and experienced less trouble.
Summary: Despite some technical problems, we were able to test interesting hypothesis for a very low cost. Currently we are working on related publications.
SETUP 2:
A demo presentation during a poster session.
Task: Gaze based interaction with images on laptop.
Equipment: Laptop PC, EyeTribe.
About ten different users joined to this demo session. I asked them to look at several images which were presented on a browser. During the demo I switched between several image processing algorithms and asked the user whether there is a noticeable effect on the visualization. Each demo took no longer than ten minutes.
After the calibration all demo sessions run without any problem. I was able to collect some task related feedback and users were impressed about this rather "mobile" eye tracking setup. I want to mention that I used a javascript based websocket, developed by Mr. Kazuhide Okamura, which can be downloaded from the following link:
https://github.com/kzokm/eyetribe-websocket
Kenan:
Sorry I do not have the expertise to answer this question.
Regards,
Pilar
May be some of the followers of this question want to know that there will be an eye Tribe pro in June (75 Hz, 199 US $).
Dear Yogesh,
I think you can. Please check out the documentation below and search for the keyword "pupil".
https://github.com/EyeTribe/documentation
Kenan
We have been working with Eye Tribe in OGAMA (software) but I couldn´t create a dynamic heatmap from video stimulus yet. Has anyone done this in OGAMA (or in another software)?
I haven't. And my eyetribe is currently not working anymore for some reason :(
I hope you can find a solution to your visualization problem Diogo Kawano .
- Badges
- Science topic
- Similar topics
- Psychology