Hi Pupil Team. I hope this message finds you well!

Here is a beginner in eye-tracking technology. I am using Pupil Core for some research on human viewing behavior. I have some doubts about some parameters of the device and hope to get your help.

1. For the WORLD CAMERA resolution choices, such as (1920, 1080) and (1280, 720), does the choice of these two resolutions affect whether I am using the surface to position the desktop monitor? Does using different resolutions require different additional operations when processing fixation data, such as position correction, etc.? Because of the need to ensure that at least two Apriltags are detected, I've found that (1920, 1080) seem to be able to fulfill this condition even when the user is very close to the screen. But I wonder if choosing (1920, 1080) has any other implications, whether there are more severe positional offsets, etc.?
2. In 02jupyter file in pupil-tutorials-master on Github, the drawing of the heatmap uses gaze_positions_on_surface_Cover.csv, why not fixation data. Is there any difference between these two methods?
3. There are two options for file size of the record, is there any difference between bigger and smaller besides CPU usage? For example, how detailed is the data, how accurate is it, etc.?
4. How should eye camera resolution and frame rate be chosen for a study of user viewing behavior? There seems to be no intuitive difference.

I appreciate your help.

Hi Pupil team! I would like to know the specific parameters of the internal human eye camera of your Pupil core product, such as its Focal length（mm）and its sensor size. Can you tell me a way to get this information?

hi..i have pupil wearable glasses and i have been collecting data earlier thro cloud.In that the gaze positions were not normalised and was in absolute values. But in offline data collected the norm_x and norm_y are giving normalised values. However for my algorithm i need absolute values. It is just scaled and reference is changed but from where do i find that scaling factor.

Hi all, may I ask what is the reason for this offset? Got two users to do this both produce this offset.

i use this

I sent the relevant recording files and processing scripts to the email address [email removed] but no response yet.😭

It does not seem to have come through. For what it's worth though, you can load your recording in Pupil Player with the surface tracker plugin enabled. If the corners of the surface do not stay "glued" to the corners of the display for the duration of the video, that would indicate some issues with marker tracking and the tips in my previous response should be reviewed

Thanks for the reminder. As well as I have re-sent a copy with a different email, please check it later. Thank you very much for your patience and help!

Hi Newbie to pupil labs core. Trying to setup the python software from https://github.com/pupil-labs/pupil. The line python -m pip install -r requirements.txt fails to complete. How do I troubleshoot the cause/error I have attached error output as a file. Thanks

Yes you can ask. I initially thought that I had to build it from source and I have subsequently installed version 3.5 as a windows package. So I can get my daughter up and running. However I would like to be able to build it from source so that we have control of the sequence of starting a recording and then trigger visual similia which is necessary for an experiment that my daughter needs to do. I will investigate if we can do what we require using API calls. I wondered if I needed to connect up the recording device beore building from source or if the build environment using the D: drive was the issue. Thanks for your response.

Hi all. What is the best calibration choreography when working with multiple screens? Does working with screens of different sizes (i.e., monitor and laptop) change effect the calibration significantly?

We have two different setups we are using. One has two 27 inch monitors next to each other, where the monitors are slightly angled in the middle. The other setup has the same two monitors, plus a small laptop off to the side.

Perfect! That is helpful. Thank you.

Hi, by any chance, did someone experience this: I'm not able to format the numbers in the duration of the fixations.csv spreadsheet, the numbers appear to be not in decimal as needed (at most one point per number), I racked my brain trying to fix this and I couldn't. I don't understand why the data is exported like this, since I used the application (Pupil Core) again, with the same spreadsheet application (Google Sheets), and the numbers were in the correct format. Please, if anyone in this community can help me, I will be forever grateful.

Hi all, I'm having problems with the calibration and validation process. Why is the accuracy and precision good after calibration? But immediately after calibration, the validation accuracy becomes poor?

Calibration settings question: which setting should I use if I printed april tags (4) and am attaching them on the rectangular surface participants will be looking at? Also, is the sample duration here the same as sampling rate. What is the sampling rate, is there a way to adjust it?

Hello, my lab has a pupil core which was purchased a while back. We thought that it is the newer version that can reach 200 Hz sampling for sampling eyetracking data, but when we test it out we're only getting around ~125-130 Hz. Is there any way to check if the model we have should be able to reach 200 Hz?

I'm attaching a photo of the eyetracking camera in case that helps.

Hello, I would like to ask how can I get a proof document of CE mark for the core version glasses?

Hi all, I have a question about post-hoc calibration. There are two reference sections here, the first is the calibration at recording and the second is the verification at recording. When I "Set from Trim Marks" in the post-hoc calibration, should I include both sections or just use the first one?

is there a python script out there that we could use to collect data from the eyetracker through psychopy?

Hi! I have a question about the fixation detector on pupil player. Though I have successfully exported data previously, I am running into the issue where the fixation detector does not gather fixations from the entire video. I am mostly struggling because I have made sure that the parameters for detecting a fixation were the same as previously exported. The only adjustment I have made to the video file was adding another surface. Is there anyway to troubleshoot this issue?

Hi, I'm receiving the following warning when running pupil capture: [WARNING] plugin: Failed to load 'pupil_labs_calib'. Reason: '[Errno 2] No such file or directory: 'C:\Users\Owner\Documents\MATLAB\PTBtransfer\NBackTaskBatteryPTB\LSL_Recording_Status.txt''. I have 'pupil_labs_calib' but this directory does not exist. How can I set the application to look in the proper directory?

Do you known of anyone that ventured into an EyeLink->Pupil data conversion tool?

Hey, why does the Enrichment creation in Pupil cloud take so long to process when running? and also what is the next step for the heatmap creation?

Good morning, I bought the Mac that you recommend for the Pupil but it doesn't have a USB port, will it work well with an adapter? Thank you!

For example, specifying the minimum stuff that a data folder must have to be properly loaded into Pupil Player.

like this

Also, In the past, I remember people saying that 2d calibration produced better results in the long run. For me, 3d calibration freezing the eye model after the participant eye rolling is good enough, so I am sticking to it.

What do you mean by raw recordings?

They might be useful if you need to look closely at data formats

Ownn, yes, I have a bunch of them

Hi! Have already a pair of Neon glasses but I'm also interested about the Core product for future development. I don't find detailed info about the difference in terms of eye tracking accuracy between these two products, as well as a documents showing differences in terms of features. Can you please help me? 🙂

hey all! I'm trying to use https://github.com/teresa-canasbajo/bdd-driveratt/tree/master/eye_tracking for saccade analysis and I don't have any prior knowledge regarding this does anyone know a better and easier way to analyze the saccade data from existing eye data? Also, is anyone working on the saccades analysis or any eye-hand movement analysis please DM I greatly appreciate your help thank you!?

I'm looking for a python example code that captures video frames using the eye camera. I have the Core and VR add-on.

When I run cv2.VideoCapture(0), I encounter the following error. [email removed] global obsensor_uvc_stream_channel.cpp:159 cv::obsensor::getStreamChannelGroup Camera index out of range

Hello all, where can I get some info on using Pupil Core with the new PsychoPy plugin? I need to know if it's supposed to automate my calibrations and recordings, or I still need to use my own code for all that?

Hello! I am trying to use pupil labs core software, using the set of pupil labs cameras that is used for tracking with the HoloLens, but in a custom configuration for a specific study (ie not mounted on the HoloLens) and only performing monocular tracking of my left eye.

When I configure the cameras such that the world view camera is mounted directly above my left eye, the world view is upside down in pupil capture. Unfortunately the camera needs to be mounted upside down in order to be aligned above my left eye.

I cannot seem to find a way to flip the world view camera in the software.

Any advice on how to do this rotation of the camera feed?

Alternatively, do I need to have the world camera positioned directly above/vertically parallel to my eye, or can it be offset?

Thanks!

My development environment is Windows with Python 3.9.13 and OpenCV-Python 4.9.0.80. In Pupil Capture, the camera video is displayed correctly. I have tested camera indices up to 10000.

While surfing the web, I came across a suggestion to uninstall the device from the Device Manager and then run cv2.VideoCapture. I tried this approach. When testing with Pupil Core, no errors occurred. However, only the front camera is displayed correctly, and the eye camera appears as a black screen.

Good morning, I bought the Mac that you recommend for the Pupil but it doesn't have a USB port, will it work well with an adapter? Thank you!

Hi @user-9973bb ! The camera can be offset and it would not matter if it is upside down, although it might be easier for you to work if is properly placed. You could write a plugin that flips the scene camera or modify the source code.

hi Team, happy new year! I am wondering if there are any existing resources that i can look into to analyse the pupilometry data and gaze data using python for data collected from multiple subjects? I have data for about 15 people for a few scenarios and I would like to compare the difference of their eye response for this different scenarios and try to summarize for all 15 subject whether they responded in a similar way.. it is quite a task but I hope there are examples I can use to pre-process and combine them. Thanks

hey all! I'm trying to use https://github.com/teresa-canasbajo/bdd-driveratt/tree/master/eye_tracking for saccade analysis and I don't have any prior knowledge regarding this does anyone know a better and easier way to analyze the saccade data from existing eye data? Also, is anyone working on the saccades analysis or any eye-hand movement analysis please DM I greatly appreciate your help thank you!?

Hi everybody, and happy new year ! I am processing some data acquired via pupil core, using pupil player. I am looking for specifics events in quite long recordings. Given that I have the timestamps of these events, I was wondering if theres any way (eg. Plugins) to jump directly to these timestamps in pupil player, as this would avoid me to have to search manually for those ? Thank you @nmt !

Hi @user-4514c3 ! In MacOS the bundles need to be started with administrator rights, have a look here.

hey, pupil labs! I'm trying to use https://github.com/teresa-canasbajo/bdd-driveratt/tree/master/eye_tracking for saccade analysis and I don't have any prior knowledge regarding this, is this the one that should be used or is there any easier or better method?

What I want to do is try cv2.VideoCapture in Python. However, I'm encountering the 'cv::obsensor::getStreamChannelGroup Camera index out of range' error. According to https://github.com/pupil-labs/pupil-docs/issues/219, it seems that I need to uninstall the Pupil Labs driver. Is this the recommended solution?

Hi Team, i started to plot the pupil diameter and I am wondering whether my dataset has a lot of noise in it. Because compared the example tutorial, it seem to still have a lot of peaks eventhough I have already filtered the data to only .95 confidence level. My chart is as shown here.

Hello Devs! I inquired long back and forgot to follow up regarding data lag issue I have been experiencing. We are running a driving simulator which has an incredibly complex architecture. Without going into the details, our driving simulator subscribes to a surface defined in the environment and uses the OnSurf boolean value to determine if the driver is looking at the specified surface. We implemented the ZMQ PUB-SUB protocol in the C++, as the project is built in unreal engine. The eye data is fetched in the driving simulator on a parallel Unreal Engine's Runnable thread. We experience a lag between the drivers actual gaze location (i.e., if they are looking at the defined surface) and the OnSurf value which we retrieve. This lag increases as our simulation trial progress. So, we are assuming that this is a computational issue. It is even more likely since the driving simulation occupies 70-80% CPU, ~50% RAM, and ~90% GPU. But we have very capable hardware, 64 GB RAM, Intel i9 12 core processor, RTX 3080 with 10 GB VRAM.

It feels like we have reached a dead end with our project because we don't know how we can optimize our software. Since there is an increase in lag, we though that reducing the buffer queue and dropping all the older messages once the new messages are received (like a queue) can be implemented. In fact, ZMQ already has functionality to do this, ZMQ_CONFLATE. However, this is not supported for multi-part messages which is how its implemented by the the pupil API. Note that in our software, we are only interested in the latest data whenever recv() is called. This goal should significantly reduce the computational issue we have (yes?)

Do you have any suggestions?

While breifly skimming through the discord message history, I see that now one has experience such a big lag (of a few seconds), only in miliseconds. Could this not be computational issue? I know I haven't even given you the details, but we just need some potential suggestions on what the next steps could be.

Hi @user-572e3e ! I assume that you run Pupil Capture on the same computer as the simulator and that you need to know whenever the driver gazes on the surface and the coordinates in realtime, is that correct?

Decrease Capture workload: - You can run Capture in a second computer and then simply subscribe to onSurf and surface coordinates in the simulator, this way all computational resources dragged by the capture would be not blocking the sim. - If you do not need fixations or other plugins in real-time, you can disable some of these plugins that you do not need in realtime.

Consuming the ZMQ: I am not sure how you implement this but here you have the Unity example and the Python reference as examples.

Optimising the simulator: The resources that you mention seem great, but it seems the simulator is devouring them, is it well optimised? I don't have much experience with UE but in Unity baking the light makes a huge difference. Btw are you on UE4 or 5?

Thank you so much for the quick response @user-d407c1 , I really appreciate it!

We are actually planning to also record and compute the cognitive workload of the user using the available physiological measures implemented by the eye-tracker such as fixations, pupil diameter, etc. So, it is not possible to disable them, unfortunately.

We are actually using modified version CARLA (https://carla.org//), which is a driving simulator used for autonomous vehicle research and is terrible in terms of optimization due to its goals of supporting diverse use cases. It is built in UE4.

Unfortunately, the Unity and Python implementation would vastly differ in UE4 because it does not natively support ZMQ and desterilizing msgpack data, so we had to use third-party plug-ins (for ZMQ and) to desterilize it, which has one SINGIFANT limitation. It only supports string data type serialization/deserialization. So, we modified zmq_tools.py to send the dictionary with string key-values pairs (very similar to JSON) for the specific surface we have defined. I know this is a downgrade in performance, but it shouldn't be significant, no?

I have a few questions: 1. Setting up a different environment for independently running pupil-eye tracker is viable for us, but how much of the latency will increase compared to using one system. 2. Do you have any implementation in C++ 3. Why does pupil eye-tracker speratly sends the topic and message (as a multipart message)? Is it becuase of the PUB-SUB model? Can the code be modified to send it as single message, or does it break the code?

Hi @user-572e3e !

We are actually planning to also record and compute the cognitive workload of the user using the available phycological measures implemented by the eye-tracker such as fixations, pupil diameter, etc. So, it is not possible to disable them, unfortunately.

You can obtain fixations as a post-hoc parameter in Pupil Player, that's why I asked if you need them in realtime.

Ohh! I see, I did not knew it.

Hmm I see, perhaps if you use a 3rd party computer for capture, you can get the ZMQ messages and transform them to something you can stream to UE (just thinking out loud).

Reg. 1, latency in this case would depend on how you connect those two computers, ie. what network you use. If you use ethernet, it should be not much.

1. No, unfortunately we do not have a C++ implementation.
2. Have a look here on how they are integrated.

You can obtain fixations as a post-hoc parameter in Pupil Player, that's why I asked if you need them in realtime.

Oh I see.

These are some great starting points, and I will have to think about this more when I am back in lab tomorrow (its actually 1 am now). I will get back to you tomorrow. Thank you so much!

Hey, I hope this isn't too simple a question, but I could really use your expertise. I'm working on a research project aiming to detect BPPV disease using pupil and head movement recorded from these glasses. Currently, i am looking to create a system where doctors can test users with the glasses and annotate the data/recordings, which will be stored in a server database. I'm wondering if there's a way to capture and record data from the front-end/client side, like in a React app. I'm a bit confused about it, and your feedback would be much appreciated.

@user-d407c1 pupil core

Thanks for confirming. I am a bit confused. Do you want the doctors to use this frontend app rather than Pupil Capture? or both?

If they are going to run Pupil Capture, nonetheless, for the calibration, are you aware of the plugin's system ? Have you considered writing a plugin with a text box that includes this info and is stored in the user_info.csv ?

That way, the doctors would only need to use Pupil Capture and then you can upload the whole recording folder.

As a side note, may I ask how you record head movement?

@user-5bac8b You can fully control Pupil Capture using our Network API. It seems a bit convoluted what you want to achieve, so perhaps you want to work over the source code.

I know that you have Pupil Core, but have you considered using Neon ? It does come with a 9dof IMU sensor, it is calibration free and to me sounds like a kind of a turnkey solution. Furthermore, you can use its realtime API to stream not only gaze or video but also the IMU data.

Hi team, i am not sure if this is within your scope of support or maybe others in this community could respond to. I am able to pre-process the pupil data and used the steps as suggested but my question now is what kind of statistical models can be used to compare the changes in pupil dilation between the participants and the different scenarios they were exposed to. So the question i am trying to answer is whether the pupil dilation was more in 1 scenario compared to others and whether this pattern exist among all participants. Thanks

Hi ! Is there someone who can explain me how to download Pupil Player ? thanks

Hi @user-71f1de ! you can download it form here

Thank you !

Hello,

I would like to obtain the coordinates of gaze on a PC monitor via the Network API. Is it necessary to use markers for this purpose? Additionally, if markers are used, is standard calibration also required?

Thank you!

Hi @user-2d96f8 ! That depends on how you plan to detect the monitor and translate the coordinates from the scene camera to the screen.

Regardless, Pupil Core requires to be calibrated to obtain accurate gaze data.

If planning to rely on the surface tracker to track the monitor position, then yes, Apriltags markers are required.

Then you will need to subscribe to the surface coordinates topic.

• If this is to control the cursor, have a look at previous community projects like https://github.com/emendir/PupilCore-CursorControl

• If this is for a experiment, I recommend that you take a look at our Psychopy integration, which simplifies this.

Thank you! I would like to select targets based on the gaze position on the screen, so it's similar to controlling a cursor. If I can display tags on the screen at all times, is it okay to use just the four tags on the screen without any physical tags?

Hi,

Could you please explain how to subscribe to the surface coordination topic from Python (or tell me the documentation)? I looked through the documentation at https://docs.pupil-labs.com/core/developer/network-api/, but couldn't find it.

I had one more question: Basically, the purpose of retrieving the real-time OnSurfvalue was to design a timer that computes how long has the participant/driver been looking at the head-up display (HUD). In our pilot study, we observed that the timer was inaccurate: the timer would reset as soon as the gaze would fall out of the HUD (as expected); but this reset frequently occurred unintentionally: as a result of the user blinking, the gaze would jitter (I don't know the correct term - I mean short term irregular movement). I would assume this is a obvious issue with gaze mapping as it could not be determined when the user blinks. However, is there a way to "smoothen" the gaze, like the commercial Tobii eye-tracker does? Is there an internal implementation/feature for it? For now, we just set a OnSurf False value threshold: the OnSurfvalue has to be false for more than 150ms (which is the average blink time) in order to count as a gaze-shift and reset the timer. Do you have any alternate reccomendation?

Hi, I would like to do a calibration, but the button is gray. Do you know how to fix it?

Hi. I have a question about the 3d eye model usage. As you may known, one can freeze the eye model. So,

1) Is it right to say that a freezed model will not compensate for frame slippage?

2) Is it right to say that you should freeze a model during calibration?

Thanks

Hi, I am thinking of buying the Pupil Core eye tracker but I have a few questions:

• What is the estimated delivery time to College Station, TX, USA
• Is it a plug-and-play device, or does it require software modifications
• Will there be technical support provided in case I encounter any issues with the hardware or software.

Hi, would you please help me add two eye trackers to a group? I have installed the plugin, but they can not find each other in the pupil groups!

Hi everyone! How to create a Plugin that inherits from the Surface_Tracker?

I am trying to update my screen tracker plugin from earlier versions.

1) Is the plugin Api supposed to support multiple inheritance?

class ScreenTracker(Plugin, Surface_Tracker)

2) Is it supposed to support single inheritance from existing plugins?

class ScreenTracker(Surface_Tracker)

Best regards,

Hi there. We have a pupil core device that we are trying to record through pupil mobile, but it seems the mobile is not detecting the glasses. I can't seem to find any troubleshooting advice online. Can anyone help?

We know it's not a hardware issue because the glasses work perfectly when wired straight into a laptop

@nmt it's Johnny Parr btw, hope all is good!

Now I will try to update my participant driven calibration with the new choreography api

Hello, I'm currently using a pupil core device on an Ubuntu system and encountering some issues during calibration with pupil capture. Despite following the instructions, the calibration marker remains stationary at the center and it's unclear which key to press for confirmation or to exit the calibration process. Could anyone offer some guidance or tips on this matter?

Hello everyone, I'm currently using a Pupil Core device on a Mac and attempting to remotely control it via the Network API. I've successfully used the API for recording and remote annotations by following the guidelines from the pupil-labs/pupil-helpers GitHub repository. However, I'm encountering difficulties with remote calibration. While I can initiate the calibration using "socket.send_string("C")", it concludes too quickly. I'd like to monitor the calibration process in real-time and conclude it with "socket.send_string("c")" once all markers are successfully calibrated. Could anyone guide me on how to track the calibration status?

By the way, in regards to calibration with the Pupil Core device, does the process only calibrate the geometric relationship between the world camera and the eye cameras, or does it also take into account the positional information of the wearer's eyes relative to the eye cameras?

This way it will be possible to calibrate only when necessary. Of course, it requires a really comfortable setup, too much movement in an uncomfortable setup would require a lot of calibrations

Pupil Mobile

Hey, I am attaching a picture of the Pupil Core that our lab has. As seen in the photo, there in a triangular mount and eye camera only for the right eye. I want to collect data for both eyes. Can someone suggest what to be done for this case?

@user-cdcab0 , is your tracking strategy available somewhere?

FWIW, the Core scene camera can detect some IR light. In the lab where I did my doctoral studies we made AprilTag markers that were backlight with IR LEDs. It worked well enough for our study

In my i5 notebook, with a discrete NVIDIA, the eye camera in running at 90 fps at 640:480

Do you know if Pupil Labs have any plans to fix the issue i opened about camera intrinsics?

Also, I think there is some error in your requirements.txt. I need to downgrade the packages module to version 22.

Also, for some reason, I am getting no error messages. The world process is crashing with my camera when running from source only.

Hello, I have pupil video recorded via pupil capture/service, but my world video was recorded via a different service (calibration marker in that wold video). How can I import and sync the world video into pupil player so I can perform calibration and gaze mapping? I've tried including the world video in the same folder with the pupil video as "world.mp4" but nothing shows up in the pupil player world window

Also, my apolgoies for the naivete, but it's not possible to get gaze measures like fixations and saccades solely from pupillary data? You need calibration to map pupil data to world positions to get those measures?

Hi there,

We are trying to compare the Tobii 4C to Pupil Labs Core in Unity to run some calibrations and short trials in 3D space, we are using hmd-eyes for this.

At the moment, we are a bit puzzled by the output data from hmd-eyes. I understand that 0,0,0 is at the centre of the camera, I assume the world camera? Or is it each eye's camera? Then, our other question is, how do we find out what units are being generated in the data files (i.e. for gaze point -0.13, -0.18, 0.93 for one eye)? I can't find anything explicitly stating - I assume the units are arbitrary and so we need to generate some points of a known location and then transform the points into the camera space or vice versa for x and y, then use two eyes to work out the z position by intersection? If you could point us in the right direction that would be a great help!

In the end, these two systems you want to compare are also different in their nature as well (fixed eye-tracker vs wearable eye-tracker). Not sure how well they will compare, but is an interesting study.

I want to detect the blinks in the Pupil Core eye videos. I'm aware that there is the built in blink detection in the pupil player software but want to explore alternative solutions. Does anyone have some advice on a robust detection method?

Hi

I am currently trying to analysis some recordings (from a pupil core headset) using pupil player. However I am experiencing a serious problem with pupil player crashing. I have repeatedly tried to run post-hoc calibration and gaze mapping but player crashes at the end or as soon as I try to do anything else and I have to force close it from task manager. When I reopen the recording after restarting player it says the mapping was completed however no gaze position appears in the recording and when I click export the exported gaze position csv file is empty. I have also found that player also crashes when I try to close the program, even if the only thing I've done is open a recording, wait a few minutes, then try to close. The only thing of note that I can thing of as a possible cause is the recordings are quite long (approx. 20 min).

Do you have any suggestions as to how I can solve this issue and prevent pupil player from continuously crashing? Or is it the case that player just does not work with long recordings (which would be a real pain)?

Edit: I have just tested out player with a much shorter 5 min recording and player still crashes when I try to close the program. The max memory use is only 50% which occurs after the crash and CPU max is around 25%. I am using Pupil player v3.5.1 on Windows.

Dear PupilLabs team,

I have been doing experiments with the PupilLabs core. Now I am trying to open the recordings in the Pupil Player, but it immediately crashes. The experiments were done by a student of mine and I am afraid that she started a recording in the software and then ran my script which should also start the recording via the API and set the pupil timestamp as it is described in this issue: https://github.com/pupil-labs/pupil/issues/1580

However - I need to find a way to access the data and get the gaze data since the experiments have already been done.

Find attached a gaze_timestamps.npy file

I hope you can help me,

Kind regards,

Cobe

[email removed] Our fixation filter does use

Hi @user-b91cdf ! After loading your recording into Player, there should be a log file being generated. Could you share it?

Thanks for following up. Do you mean the file player.log under pupil_player_settings does not get updated? Could you also share whether you are running Player from the bundle? or source? and what version are you running?

here is also the info.player.json from the capture app:

Thanks for confirming. Could you please share one of these recordings that does not load with data[at]pupil-labs.com, such that we can further investigate this?

Simply upload them to any provider of your choice (Google Drive, OneDrive, etc) and invite that email.

sure - thanks!

Hello all, I'm trying to run the pupil capture from source in Ubuntu Touch emulator (linux based system designed for phones, but I'm using emulator not the real one). I installed all the requirements, but I get the error when I run "python3 main.py capture": b"The GLFW is not initialized" However pip shows that it is installed, and I even compiled it from source from glfw repository. Do you know by any chance what else could be done to solve the issue?