Is there a way to export the frame index for the eye cameras when exporting raw data in pupil_positions.csv? I need it for one eye only. but the only frame index which is exported in the csv file is the world_index
Hi @user-4a1db0 , the data in that file are directly derived from the eye camera images, so the row number already implicitly indicates the eye camera frame index. Just note that there are two rows per eye camera index:
The 3D detection pipeline (and diameter_3d) is what most people want, so filter first on the method column to only select the 3D method before counting eye camera frames.
Thank you!
Hi does anyone have an ai that interprets data collected from tobii glasses for client use friendly
Hi Team, I have a question about the eye tracking calibration procedure as this is really important to ensure accurate results. If I have a driving simulator with a projection screen and a 200 degree field of view, which calibration method should I use? I cannot render the calibration markers within the virtual environment so I need to know if it is possible to place the markers on the screen itself and how to correctly do this. Which type of markers should be used and should I then choose the Natural Feature Calibration or Single Marker calibration. Would help a lot to know which would be the better approach. My setup will be that the eye tracker is connected to a laptop and the data recording happens on the laptop. The simulation is rendered on another computer onto the projection screen which the participant will be looking at while driving the simulator. Thanks a bunch in advance!
Hi @user-0f7e53 , is it possible to do the standard Screen Marker calibration on the laptop where Pupil Capture is running? It is not necessary to do the calibration on the same screen as your simulator.
Hi Rob, thanks for the reply. if I understand correctly, it is enough for the participant to just look at the points on the laptop to calibrate and not on the area of the projection screen? this is possible, but that will mean that they will need to first be seated on the simulator, have the laptop on their lap, use the standard marker calibration and then remove the laptop from the participant (while still connected to the eye tracker). Is this correct?
That is correct. The main thing is:
You can do a brief pilot with yourself and see how that works. If it does not work well, then just send us a message here.
okay, great, I will try this approach. Thanks
as an alternative can i also try to print the circle markers and place them on the projection screen as another possible approach in case there is a tendency for head movement or unstable laptop movement? it is sometimes difficult to expect people to sit still, just in case.
If you instruct them to keep their head still, then most people can do this for the few seconds that a calibration requires.
If you print the calibration markers, then you have to switch to Single Marker or Natural Features Choreographies, which require the person to pay attention to more instructions and have more controlled movement.
I would recommend trying the standard Screen Marker calibration first and see how that goes.
Hi everyone, could using different resolutions for the two eye cameras cause any problems with binocular gaze estimation, calibration/validation accuracy, fixation detection, or subsequent data analysis?
Hi @user-74b1c6 , the team is back to work tomorrow, but may I ask why you want to do this? It would be a rare way to use the device.
Hi Rob, thanks for getting back to me. I’m evaluating a Pupil Core setup in which the two eye cameras have different supported resolutions (192×192 for one eye and 320×240 for the other), while both can operate at the same sampling rate of 120 Hz. Before using this configuration for research data collection, I wanted to understand whether the difference in resolution itself could affect binocular gaze estimation, calibration/validation accuracy, fixation detection, or subsequent analysis. I’m mainly trying to determine whether matching the sampling rate is sufficient, or whether matching the eye-camera resolutions is also important.
Since this is a rare configuration and we are unable to provide explicit support for more DIY Pupil Core setups, I would recommend doing some pilot tests and see if you find substantial differences from using the same camera type on both sides. I do not think you should run into any issues just because of different resolutions, but I also cannot make any guarantees and this is untested.
If you can provide more info about your experimental goals and why you need these two different cameras, then I can at least clarify if a standard Pupil Core setup with the same camera on each side would actually be sufficient.
The reason for this configuration is that one of the original eye cameras of our Pupil Core headset became faulty and was replaced with another eye camera. The replacement camera is from a different camera generation, so the two eye cameras now have different hardware specifications. The replacement camera itself supports multiple resolutions. (right eye). This headset will be used for an academic research study involving screen-based tasks. We will analyze users’ visual behavior, including gaze patterns, fixation metrics, and AOI-based measures. The experiment will be conducted in a controlled environment, and calibration/validation will be performed before data collection.
The reason for this configuration is
Hi, is there a way to validate when using the pupil remote API? I can't seem to find anything online.
Hi @user-6c482a 👋 ! I assume you mean using the IPC backbone? Have a look at https://gist.github.com/mikelgg93/8820c6338a968ec2692e54172f1aff20
Hi Miguel, thank you! I'm still familiarising myself with the concepts and terms. I was wondering if there was a pupil remote command (e.g. similar to pupil_remote.send_string('C')) to initiate validation. However, I have used the IPC backbone elsewhere (to get data), so I should be able to apply this as well.
No worries, let us know if you get stuck anywhere else
Hello, I’m using the eye video exporter to export my eye video footage into 15 minute intervals.
It’s been working for all files. But for one file, I select the last 15 minutes with “relative time range to export” in general settings. I’m not sure why but the exported file has 48minutes of footage (the total time of recording is only 38minutes)
Hi @user-4a1db0 👋 ! Thanks for reporting this. An export of 48 minutes from a 38-minute recording definitely doesn’t sound right, could it be the recording contains multiple eye video files due to a disconnect?
It would be really helpful if you could create a ticket at 🛟 troubleshooting and share the recording with us (including the logs from Pupil Player), so we can take a closer look and try to understand what happened.