Hey guys!

I'm wondering if the Pupil Labs Core equipment is capable of dealing with radiation?

Another thing- the application won't let me select the "shutter priority mode" or the "auto mode" - only the "manual mode" and "aperture priority mode" are selectable. Maybe that could be related?

@nmt Thank you😃

Hi all, I am familiar with the dispersion-based algorithm used in pupil player to detect fixations post-hoc. I have been reading the code and trying to understand it. So my questions are: 1) does it take into the sample frequency of the gaze data (frame size?); 2) also, when calculating duration, is missing data considered? If there are m ore than a few missing timestamps, how is the end time of the fixation then calculated. Thanks!

Hi pupil labs team. We need to calibrate the Pupil Core at 1,5 m distance. This works perfectly when using the calibration marker that is send together with the eyetracker, which is sized for 1-2.5 m distance and has a diameter of about 6 cm. We were wondering if making the calibration marker slightly bigger (when printing the marker from the docs to DinA4 it results to a diameter of 8 cm instead of 6 cm) does affect the calibration results (accuracy and precision)? A quick test showed no difference in calibration results, do you have any experience here?

Hello, Is the screen marker calibration method only for recording gaze on the screen?

I'm a brand new user desperately trying to generate some gaze data. I'm using Mac OS. My recordings show real world video but not gaze. Is that because Pupil Capture was not started with administrative privileges?

Wow! Thank you for letting me know about the onboarding. That should really help. Question: in the onlilne documentation, it says to copy the pupil capture application to the /Applications folder. Is that the same as the Applications folder?

I am working on a Macbook Pro. Can I plug the glasses directly into it without the USB adapter?

I changed the administrative permissions as recommended and still do not get gaze data. Here is what Pupil Capture looks like - apparently no data from right eye:

Hello @nmt. I am using a plugin developed by jesseweisberg (https://github.com/jesseweisberg/pupil) to automatically detect objects from the scene camera. I have placed plugin files in the pupil capture settings folder, and when I am trying to start pupil capture software I received the following warnings. Moreover, I cannot see the object detector plugin in the plugin manager (Pupil capture application). Is it the correct procedure for loading a custom plugin, and if so, why is this warning showing? So how can I use this plugin with a pupil core headset?

eye0 - [WARNING] plugin: Failed to load 'object_detector_app'. Reason: 'cannot import name 'Visualizer_Plugin_Base'' eye1 - [WARNING] plugin: Failed to load 'object_detector_app'. Reason: 'cannot import name 'Visualizer_Plugin_Base''

hi everyone! This is my first time using core to connect to my computer(MacOS) and i have trouble with it . I am wondering where can i get some help?Thank you!

okok thank you！

Getting started

Hey folks, I'm having a little trouble with processing gaze data.

For reference, here is the structure of the gaze datum:

{
    # monocular gaze datum
    'topic': 'gaze.3d.1.',
    'confidence': 1.0,  # [0, 1]
    'norm_pos': [x, y],  # norm space, [0, 1]
    'timestamp': ts,  # time, unit: seconds

    # 3D space, unit: mm
    'gaze_normal_3d': [x, y, z],
    'eye_center_3d': [x, y, z],
    'gaze_point_3d': [x, y, z],
    'base_data': [<pupil datum>]  # list of pupil data used to calculate gaze
}

And here is the 'base_data':

{
    ### pupil datum required fields

    'id': 0,  # eye id, 0 or 1
    'topic': 'pupil.0',
    'method': '3d c++',
    'norm_pos': [0.5, 0.5],  # norm space, [0, 1]
    'diameter': 0.0,  # 2D image space, unit: pixel
    'timestamp': 535741.715303987,  # time, unit: seconds
    'confidence': 0.0,  # [0, 1]

    ### 2D model data

    # 2D ellipse of the pupil in image coordinates
    'ellipse': {  # image space, unit: pixel
        'angle': 90.0,  # unit: degrees
        'center': [320.0, 240.0],
        'axes': [0.0, 0.0],
    },

    ### 3D model data

    # Fixed to 1.0 in  pye3d v0.0.4.
    'model_confidence': 1.0,

    # pupil polar coordinates on 3D eye model. The model assumes a fixed
    # eye ball size. Therefore there is no `radius` key
    'theta': 0,
    'phi': 0,

    # 3D pupil ellipse
    'circle_3d': {  # 3D space, unit: mm
        'normal': [0.0, -0.0, 0.0],
        'radius': 0.0,
        'center': [0.0, -0.0, 0.0],
    },
    'diameter_3d': 0.0,  # 3D space, unit: mm

    # 3D eye ball sphere
    'sphere': {  # 3D space, unit: mm
        'radius': 0.0,
        'center': [0.0, -0.0, 0.0],
    },
    'projected_sphere': {  # image space, unit: pixel
        'angle': 90.0,
        'center': [0, 0],
        'axes': [0, 0],
    },
}

First, I'm curious what the field 'gaze_normal_3d' is.

Second, I am wondering how 'theta' and 'phi' for each eye are established. I.E., Would (theta, phi) = (0, 0) represent the user looking straight forward?

Thanks for any help!

I have the phone buzzing, and I get a message that says Recording Error. When I unplug and connect the wire it seems to work again, but this is pretty annoying. Also the Live Stream works 75% of the time, and does not work the rest of the time. Long time loading up the page (even with the IP address of the companion specified). Sometimes it comes up, and other times it struggles.

hi everyone! When i using core i find out that it cannot connect to pupil capture .The screen went out the picture of my eyes. Restart cannot fix the problem.Does anyone know what to do? thank you !

hi i restart the default settings and problem solved!!!

Hello! I have a Core headset and I'm creating a new plugin where I read the gaze position of the eyes. As I would like to use it on a Raspberry Pi 4b I don't want to show the windows of the eyes and the camera of the world. I do "python main.py capture --hide-ui" but sometimes the eye cameras don't open. If I don't use --hide-ui I can click on the buttons and activate "Detect eye 0" and "Detect eye 1" but I want to run it without showing/using any windows. Is there any way to program it so that it can open the cameras through the python file of the custom plugin? I tried to look for information on the Pupil website and this channel but I couldn't find it. thanks!

Good morning, can you recommend a computer that works well according to your experience with Pupil-Labs? Thank you so much

Hi @user-6eeda2! If I could ask, what's your reasoning for trying to run Core on a Raspberry Pi? It has pretty limited processing power and you'll likely not be able to do real-time gaze estimation on it.

Hi guys, I'm looking for the mouse controller on this repo: https://github.com/pupil-labs/pupil. Didn't find it, googling on internet I found this repo: https://github.com/pupil-labs/pupil-helpers with this specific library mentionned previously. At the moment of running the code it shows that "ModuleNotFoundError: No module named 'windows'", package that I already intalled via conda and pip. Any ideas of how to solve it? the current python version I have is 3.11.3

When using the surface tracker plugin, do we need to draw surfaces in the pupil capture application before starting the experiment or else, do we need to draw in the pupil player application and get fixation data onto the defined surfaces.

Hi on the right is my Pupil Capture software, but the lsl relay option is not displayed. How can I solve this? Thank you

@user-6eeda2 this is another potential (more hacky) solution: https://discord.com/channels/285728493612957698/285728493612957698/1108325281464332338

@user-6eeda2 If you run it from source you can use the flag --hide-ui https://github.com/pupil-labs/pupil#command-line-arguments, although it would only hide the window, the gui would still be there

@user-6eeda2 my understanding is you would like a stripped-down version of Capture software to reduce computational load, is that correct?

Hi May I ask what caused the problem in the video, which resulted in the loss of pupil capture

Sorry, my description may not be accurate. Thank you for your suggestion. I will give it a try

Hi @user-4bc389 ! It's me again, I've looked at the video again on a larger window and it does look like is changing (so you probably did not freeze it), apologies for the confusion, you can scratch that. And coming back to the loss of Pupil Capture, wouldn't you be referring to the pupil detection, and the moments where it is in an erroneous location? On that video, this occurs due to the eyelid occlusion preventing the pupil to be properly detected, in fact if you look at the confidence value you should see a drop. You can try to minimise this, positioning the eye cameras to capture the eye from a lower angle, providing a more clear view of the pupil at instances where the eyelid partially occludes it.

Hi!

Short question. I notice that before my recording I get a good py3D model for pupil size detection if I let my participants look around. So once I have a nice py3D model I fixate the head and ask the participants not to move their eyes during trials. I keep close track of how the py3D model changes during my experiment. Yesterday I came across a participant that had a good py3D model, but once I ran the post-hoc pupil detection in pupil player, the algorithm was struggling to stabilize to a fitting model.

I started recording when the head and eyes were already stationary. Next time, should I let participants look around after starting the recording so that the post hoc pupil detection can create a fitting py3D?

I hope this is sufficient explanation I might be very vague sorry in advance.

Cheers

Basicly, my question is: should I start the recording before or after obtaining a good py3d model if I want to use post-hoc pupil detection? I'm primarilly interested in pupillometry (fixed head, eyes don't move during trial).

When I change the color of the Vis Fixation plugin circle and dot, it remains red. Is this normal?

Hello, @user-cdcab0. I am using surface tracker plugin to capture gaze patterns using the surfaces defined in the pupil player. I am drawing multiple AOIs with the help of april tags and during this process, some AOIs are being overlapped one over the other. So, with the existing plugin, is it possible to deactivate a particular AOI and keep the other AOI active?

hi a quick question i was looking into fixation files i just saw the confidence level is recorded for gazes what about fixations?

Hi ，When I set up the eye video, I found that when my eyes look down, there is always a problem where the pupils cannot be well monitored, leading to unstable gaze points and easy drift. Common methods have been tried, but they are still easy to occur. Is there any other way to solve this problem?

Hi, I installed Pupil Capture on Ubuntu 22.04 and when I start the calibration in fullscreen mode, target display is extremely slow, making it impossible to complete the calibration. I have this problem neither without fullscreen mode nor on Windows 11 (dual boot). I have not found anything about a similar problem, has someone experienced the same issue ? Thanks

Hi, @user-6ce1ed 👋🏽 - what kind of laptop is that? Depending on your hardware, Ubuntu usually requires an extra post-installation step for graphics drivers. Are your graphics drivers installed and up to date?

Yes, some integrated graphics (or APUs as AMD likes to refer to them) may still require kernel modules/drivers. First though, can you confirm that you have pupil detection enabled and the eye cameras turned on?

Oh - it's only slow like that when you're running in fullscreen mode? That's interesting and unusual - sorry, I didn't understand that from your earlier message. It's still a hint to me that it may be a driver issue.

hi guys, i'm new to eye tracking. i'm just about to start using the pupil labs core for a study that's taking place outside. unfortunately, the sun's rays are bothering me so much that the pupil calibration isn't really working. does anyone have any ideas on how i can do something about this?

You can adjust this setting in the eye windows

that didn't do much. the detectionrate of the pupil then jumps between 0.4 and 0.7

hi! trying to control pupil core with tcp messages, but having trouble. Do I just need to have Pupil Capture open and then send messages at the ip:port set?

Hi @user-75df7c ! to programmatically control Core, have a look at our network API https://docs.pupil-labs.com/developer/core/network-api/ it uses zeromq protocol

thanks! I saw that, but I can't use python. we have a custom system to run my experiment and with it I can send tcp messages, is that not enough?

Dear Team Pupil_Labs, I am a new kid on the block here & have an enquiry about what might be the best Pupil Labs system for me to purchase. I am interested in recording portable EEG/ExG [I plan to purchase a 32 channel mBrainTrain Smarting Pro system]. As I understand it, I could use LSL to integrate the eye tracking stream from one of your systems with their EEG/ExG data acquisition. I do not have the Smarting Pro system yet - I am making enquiries & trying to work out what would work best overall in terms of eye tracking systems [Core, Invisible or Neon]. My application would be indoors - ppl walking around an art museum looking at artworks - so there will not be a huge ballistic motion component. I would like to have some future flexibility to also do some task-activated recordings - we use PsychoPy to deliver stimuli in our lab to an existing tethered EEG system. [We perform EEG data analysis using MNE-Python. I would want to record EEG/ExG data using BDF format [24 bits]]. I am not wedded to any eye tracking analysis software at this stage, but my main main issue is making sure that the data acq of neuro & eye tracking data are properly synched. I am not a beginner with EEG/ExG, but am a newbie with eye tracking. Thanks for your consideration, Aina. 🤓

Hi @user-443c5a ! That sounds really interesting, I think Neon together with our reference image mapper and our Psychopy integration can fit you perfectly, but why don't you write us at sales@pupil-labs.com and request a demo so we can ensure that's the right fit.

Hi, me and my resaerch team was doing a test with Pupil Invisible and once we did the recordings, it was giving an error and did not allow to create a project. Do you know what can be the problem?

Hello. I want to do the Camera intrinsics estimation using the narrow-angle lens. I do see the scree with the dots, but when I pres "i" on my keyboard nothing happens. When I press the circular I button in the world window, the pattern disappears. Can you please tell me how to fix this? Many thanks

Thanks. How exactly do I do this? The cursor is already on the Pupil capture window, but still nothing happens when I press "i". When I press the ciruclar "I" button in the world window, the window with the circles still disappears

Hi all, I have a quick question regarding the recorded videos from the eyetracker, the 'world' videos that have the views of the cameras combined. We would like to sync these videos with recordings made with go-pro cameras. I notice that the recorded framerate is 28.980 (mediainfo), however. Is this something that can be adjusted? And is this a variable framerate? Thanks!

thanks. now it is working. However, when we press i, we get the message "World capture 10 calibration patterns" . This appears 10 times, but nothing happens. We dont think that there is a pattern that is captured and detected.

and no data is displayed

Hi! I want to estimate the gaze position of the eyes in millimeters with respect the scene camera using the Core glasses. I don't want to use the Pupil Capture software to obtain the gaze since i don't want to display any GUI and i want to run the code in a raspberry pi. What i am doing now is to access to the Pupil Labs cameras with pyuvc and then with pye3d i obtain for each eye the center of the sphere, projected sphere, phi, theta, etc. I don't know how to estimate the gaze wrt the scene camera from the data that pye3d is given to me. I tried to look to the code of Pupil Labs/pupil but, i get lost with all the files. how can i obtain the gaze? Thank you in advance!

Hey guys! Do you know what kind of infrared lights are there on the cam for recognize pupils? and what wavelenght spectrum do they have?

Hello, I am having an issue with the export process in pupil player. I have to repeat the process many times so that I can get the surfaces folder and the files inside it. I am worried that even after getting them (after many tries) some data are not exported from surfaces. What might be the issue?

you mean when downloading the file on the player gui?

Hi teams. I just a question about data filtering. Now I am trying to analyze the gaze data. For each gaze data, I got a value, like “on_surface_confidence” which present the probability of the existing of corresponding gaze. I wonder if there is an official threshold for this value? For example, I should only keep the data that has more then 0.8 confidence. Or I need to decide it myself? Thanks for considering my question!

Hello... any tips on how to make the process of adjusting the eye trakcer and calibration easier it is taking me a long time to do so for each participant i am using the single marker because i need a wider area to be covered on the screens but i always get bad calibration and sometimes it takes me more than 15 minutes to adjust it. especially the eye camera part i always have to adjust it several times to fit the user and always getting bad confidence

is physical markers on a white board better than single one?

Thank you for your management. Could you please answer the following questions? Below are questions for Pupil Core.

Q1. I have a csv file named "blinks" in Pupil Core. As shown in the attached image (left side), line F is marked as index. What does this index mean?

Q2 In the csv named "blinks," is there an index for the blink fold point (the point where the eye is completely closed)? The blink fold point is the point where the eye is completely closed between the start and the end of a blink, as shown in the right figure in the attached image.

Q3 In the csv named "blinks", there are start_frame and end_frame. For example, which of the following is start_frame?　(1) the frame at the moment the eyelid closes (i.e., the blink is recognized from the eyelid border), or (2) the frame at the moment the pupil begins to be hidden by the eyelid.

Q4 Which of the following is the blink recognition method?　(1) Recognizing the blink from the eyelid border. (2) Recognizing the blink from whether the pupil is hidden or not.

Q5 Does the frame at start_time in line B of the csv mean the start_frame in line E of the csv?　The two values did not exactly match when these correlations were calculated.

Hi @user-5c56d0 👋 ! Pupil Core has two ways to obtain blink events depending on whether your require them in realtime or post-hoc. Most of the questions you have are already answered in the documentation. Which you can find here: For Capture, https://docs.pupil-labs.com/core/software/pupil-capture/#blink-detector For Player, https://docs.pupil-labs.com/core/software/pupil-player/#blink-detector

I would also recommend reading our best practices about blink detection with Core https://docs.pupil-labs.com/core/best-practices/#blink-detector-thresholds

But generally speaking, the blink detector uses the confidence in the pupil detection, and a filter over it (here, is the posthoc filter https://github.com/pupil-labs/pupil/blob/eb8c2324f3fd558858ce33f3816972d93e02fcc6/pupil_src/shared_modules/blink_detection.py#L360). On a blink event, there is normally a sudden drop in the confidence, followed by a quick recovery which is used as a signal.

• Q1: The index, as noted in the docs, is the median frame of the world camera on that blink event.
• Q2: We do not provide PERCLOS, but you can work around by defining a proper threshold and using the median of the blink event.
• Q3: Please have a look at https://docs.pupil-labs.com/core/software/pupil-player/#blink-detector as all fields have a description.
• Q4: See my answer above on how the blink detection works in Pupil Core.
• Q5: Start time refers to the start of the blink event, while the start_frame_index is the world frame (scene camera) at which the blink event starts.

As described previously, the blink detector uses the confidence in the pupil detection, this signal is convolved with a filter and the response is named "activity".

Assuming a good pupil detection, on a blink, as the eyelid closes and occludes the pupil, the confidence in the detection of the pupil rapidly decreases and the "acitivity" signal increases.

Here comes at play, the onset threshold, which is the value at which the activity response signal starts peaking, or in some sort of way, the confidence in the detection of the pupil starts decreasing. You can define that threshold in the settings, then, as the confidence in the detection of the pupil starts increasing, the activity signal falls, and there you have the offset threshold, which determines how much it should fall to be considered end of a blink.

Although the activity response its directly linked to the pupil detection confidence and highly correlated with the openness of the eyelid, this is not a 1-1 relationship and therefore there is no fold point. You can highly increase the onset value, and lower the offset one to filter blink events where the eyelid may have not been fully closed, but again nothing ensures that the confidence totally decreased by the fully closeness of the eye lid.

The index, is the median of the world frame indexes that appear between the onset event and the offset. In other words, the start of the blink event as per the rise in the activity signal and the end of it by the signal falling under the offset threshold.

I have tried the standard screen calibration it is better than the single physical marker i discovered that there is an issue with the eye tracker so will send an email to the team about that because i prefer the single marker since i have large area to cover. I have another question in the gazes output how can we differentiate between unique gazes is it the world_index?

Hello, I broke the connector on my pupil core headset with the eye0 camera. I would like to resolder it but I could not find the reference of the connector. Does someone know it ?

Hey @user-eb6164! In gaze_positions.csv, each row represents a unique gaze datum. The world_index tracks the scene cam sampling rate, so you'll find multiple gaze datums per world index.

On the pupil player is there a way to chop up the recording/trim the recording down?

Thanks 🙂

Hi there! I am starting a new project using Pupil Core and am excited to learn more about this eye tracker! I have been playing with pupil player and the sample downloadable data, and had three quick questions: 1) what are the units for gaze_point_3d_x? 2) should gaze_point_3d_x basically be the same as norm_pos_x (both from gaze_positions.csv) except the norm is scaled? 3) if I look at something in the real physical world, keep my eyes there, and then pitch my head down a little bit, what would that look like in the gaze_positions.csv file? Maybe another way to put it is if we get eyes-in-head or eyes-in-real-world at the end? Thank you very very very much for your time!!

Hi! I am working on a computer vision project with the pupil core. For my work, I am trying to record a few clips with the world camera (for analysis) but for some reason, the recording is being saved at a lower resolution than what I have put in the settings. Is this a common problem? Can anyone who has faced this issue pls help me?

hello! I have a problem importing files from Pupil Labs to iMotions, previously I did it by downloading Pupil Player Data but it is no loger possible to download it from Puppils Lab Cloud, can you tell me how to import files to iMotions now? There have been some changes in the software and it haven't been changed in the help center yet....

exactly

Thanks for clarifying that - Let's move this conversation to the 🕶 invisible channel. I've replied to your question: https://discord.com/channels/285728493612957698/633564003846717444/1121411569608298547

sure, thanks

Dear sir. The Pupil core eye camera is currently at 120HZ, can you tell me how to make it 200HZ?

Hi, I want to calculate the horizontal pupil position/angle relative to the head/world camera origin in degrees. I'm streaming my data through LSL, which records norm_pos_x/y and gaze_point_3d_x/y/z. At the moment I'm calculating the horizontal angle (psi) in spherical coordinates like in this tutorial: https://github.com/pupil-labs/pupil-tutorials/blob/master/05_visualize_gaze_velocity.ipynb (psi = np.arctan2(gaze_point_3d_z, gaze_point_3d_x)). I was now wondering, if I should use norm_pos_x/y instead or if my calculation is right?

I think what you have is correct. You wouldn't be able to calculate psi from x/y image coordinates because you have no depth component there.

Also, if you do end up working with those normalized coordinates for something else, keep in mind that they are normalized to the scene image resolution. Because the image width and height are different, the x and y normalized values have different scales

Dear sir. The Pupil core eye camera is currently at 120HZ, can you tell me how to make it 200HZ?

Hi @user-5c56d0 ! You should be able to change this setting in the eye camera window under video source

Is it possible to purchase an HMD add-on camera that supports FHD resolution? I heard that it was available for sale a few years ago.

Hi,

For our project we want to know when someone is looking at a particular surface based on a person's gaze data using the surface tracker feature from PupilLabs Core. However, the timestamps do not seem to match with what we expected. The timestamps of gaze.3d.0._on_surface are lagging behind the timestamp of the world video frame. Does anyone know if it is supposed to be like this, if not is there a quick way to fix this problem?

The data was logged using a hdf5 logger.

Hey folks, I'm running pupil capture from source on Linux. The software opens correctly and I can see all the camera feeds. But when I try to perform calibration, the calibration screen freezes, and I get the error:

(venv) [email removed] python3 main.py capture
[14:08:37] WARNING  eye0 - uvc: Could not set Value. 'Backlight Compensation'.                                                                                                                          uvc_backend.py:424
           WARNING  eye1 - uvc: Could not set Value. 'Backlight Compensation'.                                                                                                                          uvc_backend.py:424
[14:09:31] INFO     world - calibration_choreography.base_plugin: Starting  Calibration                                                                                                                 base_plugin.py:537
open(): No such file or directory
[14:09:34] INFO     world - calibration_choreography.base_plugin: Stopping  Calibration                                                                                                                 base_plugin.py:574
open(): No such file or directory
           ERROR    world - gaze_mapping.gazer_base: Calibration Failed!                                                                                                                                 gazer_base.py:203
           ERROR    world - gaze_mapping.gazer_base: Not sufficient reference data available.                                                                                                            gazer_base.py:297
           WARNING  world - plugin: Plugin Gazer3D failed to initialize

I don't think the problem is 'Gazer3D' failing to initialize. I'm pretty sure it's whatever's causing the "open(): no such file or directory" message. I looked at the source code in base_plugin.py and found no calls to an 'open()' method, so I'm quite confused.

Any help would be so appreciated! In return, I can keep you guys updated on running the pupil software on a single board computer 😄

Hello, can we get the gaze in Pixels of the World camera? in real time via the network api, Also what is the 0 and 1 limit of the normalized gaze values? thanks

I'm not sure about getting the gaze in pixels of the world camera. I'm pretty sure the pink dot on the screen is actually using the norm_pos parameter -- which then I suppose could be mapped to the size of the world camera

Hi Kin, thanks for the reply , the normalized values chage between 0.4 and 0.6 even if the pink circle moves from one side of the image to the other, do I need to scale the values?

That's interesting. Maybe my understanding of the normalized values is incorrect. You are looking at the norm_pos value, correct?

Exactly,

Here is the message,

https://youtu.be/h___c_abjx8

The plot in this video is norm_pos. I am re-normalizing to -1,1 on both axes, but I am getting values between the full range of 0 and 1 before normalizing

I'm not sure why your values only between 0.4 and 0.6 😦

I am using norm_pos of the gaze data after 3D calibration, is it will be helpful to try 2D calibration?

Also, what are the units of the gaze_point_3d? centimeters?

I used 3d calibration too. Not sure about the units

We are using the physical marker calibration (not the screen),

It's probably not actually frozen. Do you see one of the calibration markers on the screen? If you hit escape, does it close the calibration?

When you start calibration, the program is trying to collect data from both the world scene camera (specifically, the location of the calibration marker in the image) and from pupil detection. Those two pieces of the puzzle are both necessary together - if either of them are missing, then the calibration will not move forward (which may have give you the impression that it's frozen).

So your plugin: Plugin Gazer3D failed to initialize probably actually is the culprit here

@user-cdcab0 I've been looking through the source code to try to find where the error is coming from, and I can't find where the plugin: Plugin Gazer3D failed to initialize logging statement is coming from. I also looked for FileNotFound exceptions that could be logging the 'open(): No such file or directory' messages, but I couldn't find any that would create that message. I also looked for general except exception as e messages, but none seemed relevant.

I can't even find the words 'failed to initialize' anywhere in the repo...

I've also searched all of the issues and discussions for the same error message, and none of them are related to my issue of the calibration screen freezing.

Thanks again for any help, by the way!

This https://discord.com/channels/285728493612957698/285728493612957698/1100792975132471326 and this https://discord.com/channels/285728493612957698/285728493612957698/1098517309552869386 appear to be the same issue. I read through the responses to those. All three camera feeds are working and so is pupil detection and the 3d eye model.

@user-1c31f4 - let's start with the basics. 1) Have you made any changes to the source (git status, git diff, etc to check)? 2) Have you tried clearing your capture_settings folder and running with all the defaults? Note that when you're running from source, settings are loaded from the capture_settings folder in the source tree

I'm going to take a look and make sure I haven't made any changes when I'm back at my lab. But, this evening, I found something interesting: the issue only occurs when calibration is run in full screen mode. And, when holding down alt tab (cycling through all the open windows), the program un-freezes. The same problem occurs with the rate at which the pupil glasses sends packets via the network api -- everything slows to a halt unless I hold alt tab. Pretty bewildering. Must be something to do with the focused window -- possibly a performance setting that throttles background windows. Or, maybe when the window is selected with alt-tab, a refresh is forced -- so by holding alt tab, a refresh is forced at a rapid rate (making it appear to run smoothly).

I'm going to dig into it a bit more tomorow, and now I have a good starting point. Thank you so much for your help, @user-cdcab0 ! It's reassuring that I can move on from the 'no such file' message and the 'plugin3d failed to initialize' message. Whatever is going on is definitely tied to the 'focused window' thing I mentioned in my previous message.

I should mention -- I am running the pupil software on a single board computer, the Lattepanda 3 delta. It runs fantastically smooth (except for this calibration issue), much faster than my 2020 MacBook Air. This device is likely the best solution for users attempting to use a raspberry pi, as the Lattepanda 3 Delta is only marginally larger. It's also very easy to get the pupil software set up on it because its x86, not ARM, so running from source is not even necessary, and all dependencies worked exactly as expected. The real-time pupil detection and gaze calculations work perfectly. (Just putting this here for others info ☺️)

I also restarted with default settings, also no change

Made sure I had no changes with git, no change

Hello! I've just installed the pupil capture software and plugged in the device for the first time. The application has loaded but no video is coming from the device, just just a blank screen. Basically, I'm stuck at step 2 on the "getting started" steps. Any advice would be appreciated!

Hello! Would it be possible to record the eye images with a constant frame rate through Pupil Capture?

For example by using a Request-Response method instead of the current method. If this would be possible how high would the maximum frame rate be for 192x192 images? (Currently 200fps for Vive pro add-on)

See https://docs.pupil-labs.com/developer/core/network-api/ if you are not yet familiar with the network API 🙂

I don't yet have an intimate understanding of everything that happens on that event loop. It's possible that even if the uvc video stream callback has more consistent timings that other activity in the loop might negate the effect. I think that's unlikely, but not impossible, and it would be disappointing to put in that work only to find no benefit

Hi,

Please may I know if it's possible to trim a recording? I accidentally let Pupil Core continue recording for an additional ~10 minutes and I'd like to trim the recording before analysis. The total duration of the recording is now 30 mins which is too long and would probably crash Pupil Player

thanks!

@user-d99595 Replying to your message from the Neon channel https://discord.com/channels/285728493612957698/1047111711230009405/1123913766594170941 I'd suggest using the Network API (https://docs.pupil-labs.com/developer/core/network-api/#network-api) to programmatically send an annotation each time the participant starts a new trial on Psychopy. We offer also a Psychopy integration for Pupil Core (https://psychopy.org/api/iohub/device/eyetracker_interface/PupilLabs_Core_Implementation_Notes.html)

Thanks @user-c2d375 I'll have a look.

So far I've only run pupil core capture from source (V3.6.6), naively assuming it would be more performant with Python 3.11 etc. It usually occupies around 65-75% CPU on an 8core-16HT Core i7 Dell workstation (each eye is ~35% and world is ~2%). Much to my surprise when I downloaded the old (2021) precompiled binary V3.5 it occupies around 16% CPU (~7% for each eye capture and 2% for world capture). This is on Ubuntu 22.04. I tried to make sure the same plugins and camera FPS was used by both (I only have 2 default plugins enabled). Is this expected? EDIT: CPU measured using system monitor which is averaged across all cores (100% = all cores at 100%), pupil capture CPU values (around 600 for source per eye) are much higher as they are per core?)