Happy New Year everyone! Im currently trying to build pupil labs from source on an arm64 linux device. After a bit of stumbling i managed to install all python libraries as well as requirements. When starting pupil capture however, I get the following error message: File "/pupil/pupil_src/launchables/world.py", line 537, in world main_window = glfw.create_window( glfw.GLFWError: (65542) b'GLX: No GLXFBConfigs returned'I already tried to build glfw as well as the related python bindings from source. The weird thing is that with the pyglfw sample application I have no problems to initialize a glfw window. Also the pyglui sample runs without problems (so i dont think its graphics driver related). Does anyone know what else I could do to solve my problem? Thanks in advance!

Hi @papr. Recently, I studied the calibration routine of the software, and I have some questions about it. 1. There are two "Gazer" classes. Since we are using the HMD streaming to provide images of all 3 cameras to the Pupil Capture, is our gaze calibration done by the GazerHMD3D class? 2. Also, for post-hoc gaze calibration through Pupil Player, is it done through the GazerHMD3D or the Gazer3D? 3. In GazerHMD3D, the eye translation with respect to the world camera is supplied externally. Could you please point out which class/method supply these parameters? Also, can we manually set the eye translation parameter? 4. For both gaze calibration routine, the reference data are essential. Could you please point out which class/method provide the reference data?

By default, we raise all glfw warnings as errors. You can add your error code (65542) to this whitelist of ignored error codes https://github.com/pupil-labs/pupil/blob/master/pupil_src/shared_modules/gl_utils/utils.py#L385-L389

We are using the Pupil Lab's eye cameras and world camera, but we modified the setup a little bit such that the original eye translation hardcoded in calibrate_3d.py is inaccurate. Given that, I think I should just run from the source and modify the hardcoded eye translation in calibrate_3d.py according to our setup, right?

Hi, can we use Reference Image Marker with Pupil Core? if yes, than how we can generate reference image marker? Thank you

Hi, is there any way to trim or split a surface post-hoc so that multiple heat maps may be made for it? Or a way to merge identical-sized surfaces?

Thanks! I made a surface representing an area of a face (like "right eye") and would like to be able to compare fixation distributions between other eyes on other faces. However, as of now all eyes are treated as the same surface. However, each eye appears for only 2 seconds at a time, so would there be a way to trim by time?

All are on a screen

Yes--both faces and the apriltags

Then you could display different apriltags with every face. This way you can define a dedicated right eye surface for every face.

hello. Is there an official channel to make requests for Pupil Player?

HI everyone, is it possible to set a ROI for the auto exposure mode ?

hi,when i pip install -r requirements.txt, it happens follow problem

my computer version is win10,and python version is 3.9

how to save it?

@user-7904e8 hi. To run from source on Windows you need Python 3.6. I recommend running the bundled application. You can get it here: https://github.com/pupil-labs/pupil/releases/latest/#user-content-downloads Nearly every functionality can be added via plugins to the bundled application. So there are only very few use cases were running from source is necessary.

hello everyone, I want to know, is the Pupil labs core suitable for people who wear glasses?

Hello! Does anyone have a solution for our eye tracking data not uploading to the cloud? The data on the app is sitting and trying to upload but will not move past 0%. We have tried changing our wifi connection, restarting the phone and app, neither have worked.

hi i have a quick question, I have built the DIY pupil core. I cant seem to get it to select the cameras. I have verifed that in the windows camera app they both show up. I tried running the pupil capture as admin and it showed several cameras but when i try to select any of them it says the camera is already in use. How should i procede?

there is the log from the core

I also have tried uninstalling the device in the device manager

I tried to use a VM running Linux as well, in the VM the pupil capture software picked up the cameras (but had some issues with capturing the output).

is there a way to manual load the drivers?

I have attempted on both a windows surface pro 7 and a custom desktop both running windows 11

@user-5135bb Hey! Pupil Capture only installs drivers for Pupil Labs cameras by default. To manually install drivers for third-party cameras manually, please follow steps 1-7 of these instructions https://github.com/pupil-labs/pyuvc/blob/master/WINDOWS_USER.md

ahhhh okay perfect i will give that a try

Fantasitc it is working great!! thanks for the help

Will do!

Hello, I have errors when using the eyetracker. I just received it and I have a computer with windows10. I have installed "pupil_v3.5-1-g1cdbe38_windows_x64.msi", and when I connect the core to the computer via USB, and I launch pupil-capture (v3.5.1), it only launches a black window, and nothing happens. I don't see the interface of the software appearing... even after an hour of time. I have uninstalled, reinstalled, restarted, and still the same thing happens.... Can you help me?

Hi, i have only one windows, the commande line prompt.

For context: The bundled application is compiled for a specific CPU architecture and using a different one causes the software not to run properly.

i have an alienware laptop. Intel Core i7-7700HQ CPU 2.80 GHz, 16 Go RAM, x64

My pc is running under Windows 10 Professionnel. I try to launch "pupil_capture.exe" from the commande ilne prompt. It launch the same black window, like the double-cilck. No error messages on the commande line prompt. Do i need a specific python version on my computer?

the capture.log: 2022-01-13 14:58:48,976 - MainProcess - [DEBUG] os_utils: Disabling idle sleep not supported on this OS version. 2022-01-13 15:54:59,801 - world - [DEBUG] launchables.world: Application Version: 3.5.1 2022-01-13 15:54:59,811 - world - [DEBUG] launchables.world: System Info: User: p.labedan, Platform: Windows, Machine: port-labedan, Release: 10, Version: 10.0.18362 2022-01-13 15:54:59,811 - world - [DEBUG] launchables.world: Debug flag: False

Every time the application launches, these logs are overwritten. So there seems to be something different between yesterday, when the application ran at least partially, and now where it is not even overwriting the logs anymore.

So the application launches, but seems to get stuck before the window is created. Could you please restart the computer (to ensure that there are no dangling Pupil Capture processes running somewhere), delete the user_settings_* and capture.log files in the pupil_capture_settings folder and then try launching the app. Please for a minute and share the newly generated capture.log file with us in case the issue was not resolved.

Indeed. Something was blocking the application for over an hour. Is it possible that you are running an anti-virus or similar administrative software?

@user-6cbd8b if no camera is detected, then it is likely that the drivers were not installed correctly. Check out https://docs.pupil-labs.com/core/software/pupil-capture/#troubleshooting

Mac and windows which is better for eye tracking?

Hello

Please I have a question.

In the video, how do you know the 0,1 X and Y coordinate of the screen so we can convert to the actual screen size in centimeter? The 0, and 1 are off the screen

Hello everyone

I want to start using Mac, would all my eye tracking file on window open on Mac ?

Hello people, In my setup I want to do a Natural Feature Calibration from inside a car. I have placed different markers at different distances and angles. My original plan was to not allow any head movement during the calibration, as I am then calibrating to the largest FoV possible. However, during validation I wanted to allow the head movements to cover the normal glances and head movements like during driving.

So I get an accuracy of 2.58+/-0.09 after calibration and 1.726666667° after validation.

However, I noticed when I do not instruct the subject how to hold his head, I get a much higher accuracy. (Normal Gaze-behavior)

After calibration: 1.03+0.08 °. After validation: 1.3144444°.

What do you recommend ?

Allow head movements in calibration and validation Do not allow head movements in both cases or a mix ?

Problem: If I allow head movements (no instructions to subject) , I calibrate to a smaller FoV and can't make any statement about the accuracy at large eye angles, but accuracy is much better and calibrated to normal gaze patterns.

Cheers, Cobe

0 and 1 of the surface coordinates are based on the surface definition. Where ever the top right surface corner is, it corresponds to (1,1) and the bottom left to (0,0)

Best practice would be to cover all gaze angles during calibration that you might expect, and want to capture, in your experiment. For example, if stimuli appear in the periphery of vision (e.g. in the wing mirror), the driver may glance at the mirror without moving their head. In such a case, it would be worth having the system calibrated for that.

You could subsequently validate with or without head movement: With - provides accuracy during more typical gaze (majority of eye movements) Without - provides accuracy at larger angles (minority of eye movements, but possible, e.g. glancing at the mirror)

Note that even if accuracy is reduced at larger viewing angles when compared to more typical gaze, ~2.5 degrees is likely sufficient to identify the object being gazed at. This of course depends on the size of the object, but it certainly sounds reasonable in a driving situation.

Alternatively, if you aren't interested in those glances to more extreme angles, and most of the experiment contains gaze accompanied by head movements, then maybe calibrating within the smaller FoV is sufficient.

Like I need to define a big surface for the whole stimulus right ? The size of the stimulus is from 0.22 to 0.78 and converting to actual distance in centimeter is difficult because we don't know where 0,0 and 1,1 are

Hello papr, i tried what you suggest, but it's the same result (https://docs.pupil-labs.com/core/software/pupil-capture/#troubleshooting). Tha cam is detected by the system, but not used by pupil_capture.exe...

This helped... Thanks so much

After the test you recommended me, here's the device manager when i connect the eyetracker on the usb. I'm still waiting since approximatively one hour that the pupil_capture.exe responds... because i remerber that the cam whare detected differently than the on the jpg.

Hi! I have a question about a problem that I started having yesterday and I haven't found a solution. When I open the files of my experiment in the pupil player I can no longer see the video of the world camera, I have been working on this file without problems for a week and suddenly yesterday it stopped working. I can find the video in the folder but even re-importing it into the Pupil Player I still see all gray.

Hi, I am having issues connecting the pupil camera to the pupil capture software. Every time I try to connect it, grey screens appear. I tried using the device manager and uninstalling the drivers and restarting the computer with no luck. I also logged in as an administrator and ran into the same issue. Please let me know if you have any recommendations in order for my device to recognize my pupil camera.

Hi, I am sorry to hear that. It sounds like your video file got corrupted and Player has difficulties opening it. If the video file playable in every other media player, please try shutting down Player, deleting the world_lookup.npy file and restarting Player.

Hi, is there any apps needed to be installed on my pc so I can transfer my videos from companion device?!

The 3 cameras are listed under the camera category and libusbK USB devices category.

As long as we created a surface before exporting the data ?

Hello, how should I read the values of the timestamp column? How do I get from the values to seconds? The same applies to the diameter_3D column? Which unit is used? The values differ very much from each other. See example:

Hello. The latest pupil capture 3.5 doesn't detect the cameras with the latest Pupil Core in Montery despite running with root privilege. "Could not connect to the device'. Has anyone able to run it successfully ? Thanks a lot !

hi all. I serached through docs and videos, but can't see the info on "best practices" of matching the worldview camera to real "worldview" of user

Could you give an example of what you try to achieve?

using monitor as reference leads user to shift head to match the worldview from monitor. - liftng worldview up to ceiling and asking user to comfortably watch the center of monitor and without shifting his head realign worldview camera to match center of worldview image kinda works. but then we get into situation that user can move the object and place it too low in worldview image during experiment.

Pupil Capture assumes that the relationship between scene and eye cameras is fixed during and after calibration. Adjusting the scene camera breaks this relationship and requires recalibration.

it leads to problem that user calibrates to ~50cm distance, but then can move object closer to 30-40cm distance, or move it lower than worldview was set.

with calibration we have no problem - best practice we found is to calibrate using monitor so that monitor is set to same distance as objects the user interacts with

problems are with aligning worldview - before calibration

You can always use the 1920x1080 resolution which has a huge field of view.

hm.

so using lower resolution is not scaling but like digital zoom - cropping smaller part of sensor - and making picture "zoom in" ?

but with huge FOV we "lose" ability to do good calibration with small monitor - as you can't put it "in your face" for user. and if you set it to 50+cm you occupy very small part of picture and calibration results are not very good.

I see that we need to experiment more and find better methd for our usage scenario.

Hi @papr thanks a lot for getting back quickly. Apparently, the problem is in the USB adapter as the new Mac has only USB C. I tried a bunch of adapters. Some work with the new Pupil Core with single camera while the others work with new Pupil core with two cameras. Thanks a lot once again !

I have 2 questions on worldview - 1 why it is not sed directly in the middle of forehead? why it's offset? it causes parallax if calibration is done on longer distance than objects in experiment. 2 - where can I read more on USB-C option for external cameras-sensors

@mpk might be able to comment on your first question.

@user-a07d9f the USB-C world cam option is used for prototyping. You will need to "bring your own" sensor and will need to write your own capture backend. Some researchers have used this for adding a depth sensor (previously using Realsense sensors).

ok - I'll go with fullhd for now.

By the way - is it possible to have reduced resolution but with same FOV as 19201080? like 960540 to have direct 2x downscale?

hm and how zoom with pan and all this works then? is it sw processed on pc ?

We tried that with the Intel RealSense camera. But when Intel dropped support for the R200 from one day to the other, we were forced to do so, too. And this is not the user experience we want for our customers. The usb-c mount is a trade-off where we can provide a flexible hardware and software platform without depending on camera manufactures.

Ok I'll go test all this now. We worked with old version of Pupil Capture all the time before (2.3) and I wonder if switching it to 1920-1080 will make cpu pload much greater. are there any hw encoding features present in newer releases?

That makes me sad. "so much HW and so little acceleration"

ok - I'll go check the system recommendations for CPU ram storage and also for Mobile setup.

Hello all, is it correct to say that surfaces created with the surface tracker are defined in the undistorted view of the world camera, rather than the distorted view? I am attempting to use surfaces to track gaze on a curved computer monitor (unless this is not recommended) and am trying to address similar issues as this user was https://github.com/pupil-labs/pupil/issues/1930

Thanks for the reply - so in the gaze_positions_on_surface file generated by the surface tracker export, on_surf events refer to times when the gaze mapped in the undistorted space is within the surface as defined in the undistorted coordinates?

Hi Guys, i want to visualize the gaze angles in horizontal degrees and vertical degrees (not with the angles theta and phi defined by the eye coordinate system). To make it more intuitive. So looking straight would result in 0°/0°. My current approach is to use the norm_x_pos and norm_y_pos and scale it with respect to the FOV of the world camera and a defined Mid point. Is it possible to do it this way ? I am wondering whether the linear scaling is ok.

I am doing it this way because i want to compare typical gaze angles with my targets angles during the natural Feature calibration. If I subscribe to the calibration data via zmq i get the ref_list of the targets in norm_pos_x and norm_pos_y as well.

thanks, Cobe

Ok i clearly misunderstood the desciprtion given here: "looking directly into the eye camera: (x=0, y=0, z=-1) (cartesian) phi=-π/2, theta=π/2 (spherical) looking up: decreasing y vector component; increasing theta looking left: increasing x vector component; increasing phi"

What do you mean by "actual gaze_point_3d"? Should i e.g. place a Marker straight 0°/0° H/V -> measure the vector and subtract it from the other gaze_point_3d data ? Then i would plot everything as follows:

For better understanding, I just recorded a video only looking horizontically: -> Theta = vertical angle.

Good morning from here

I know I have asked this question before very long time ago. I so sorry if it's kind of boring.

I want to know the importance of the validation steps. How do you know the accuracy has dismissed due to slippage? And why the validation after calibration? Do we take the accuracy values for each participant or trial?

Yes ok, The generated plot is based on gaze data (gaze_point_3d x ,gaze_point_3d x, gaze_point_3d x), i exported via the pupil player. -> gaze_positions.csv . The plot is generated by the pupil-tutorial 5 - Analyzing Gaze Velocity. So the next step would be as I described here: "Should i e.g. place a Marker straight 0°/0° H/V -> measure the vector and subtract it from the other gaze_point_3d data ?" The previous norm_pos_x/y are also based on gaze data. I did not analysed pupil data before. my fault sry

Hi - Would it be possible to use the functions to process the eye tracking data in real time on a local processor? I am looking into a project where the data would need to collected and processed on moving vehicle and then the world camera with gaze position imposed is sent wirelessly back to a remote location.

Hi, yes, this can be made possible but requires fairly high processing power from the local unit.

Something around an Intel Core i5 with >2.8. GHz? Giving an exact number is difficult. Alternatively, you could also use something low-powered, e.g. Raspberry PI, and stream scene and eye videos to a computer with more processing power and running Pupil Capture.

Hi, does Pupil Labs have multi glint eye tracking systems?

I will have a look on Monday 👍

Hi there! I am setting up my lab and would like to incorporate eye tracking. I have a question regarding my setup and compatibility with pupil labs

My existing PC has a custom-made software (using LabView compiler) where the subject must track a moving dot. Would the pupil glasses allow interface with a 3rd party stimulus presentation software?

It looks like Pupil Invisible would do this. I am wondering how I would sync up the eye tracking data with the stimulus and other hand movement data I am collecting.

Hi @user-97fc9e 👋. Both Pupil Core and Pupil Invisible have Network APIs that offer flexible approaches to send/receive trigger events and synchronise with other devices (e.g. https://docs.pupil-labs.com/developer/core/network-api/#network-api). We also maintain Lab Streaming Layer (LSL) relays for publishing data that enables unified/time-synchronised collection of measurements with other LSL supported devices (e.g. https://github.com/labstreaminglayer/App-PupilLabs/tree/master/pupil_capture). So that I can make a concrete product recommendation, it would be helpful to learn more about your moving dot paradigm (and other gaze tasks) – what sort of eye tracking metrics are you looking to extract?

Hello all.

Are there any ways to troubleshoot the player 2.3.0 startup failure on win10?

Sorry to say that - but there is no pupil_player_settings folder there

Thank you - this resolved the issue. I wanted to ask if there are faqs and troubleshooting guides anywhere in docs section.

By the way - we have several cores with different revisions. One has eye cam connectors without plastic thingy which looks like a capon it and we had to repair the wires seveeal times. I wonder if it is possible to order those caps?

Thank you

Hi Neil, I'd like to measure pupil dilation, duration of gaze fixation, and saccadic movements (trajectory, distance, speed) as the subjects move their eyes along the line. The dot starts at the center of the screen and moves along a straight line in one of 28 directions.

Hi @papr . I want to manually save and load 3D eye models, so I tried to dump self.detector into a file in the pye3d_plugin, but I got a RuntimeError: SynchronizedArray objects should only be shared between processes through inheritance. I also tried creating a method in detector_3d.py (the detector class in pye3d package) to access the 3d eye model, but I still get the same RuntimeError. It seems like an issue related to multiprocessing, but I cannot find anything related to multiprocessing in pye3d class (in file detector_3d.py) or pye3d_plugin.py. Could you please point out how I should solve this issue? Here is my modified pye3d_plugin.py:

Hi @papr , got a quick question regarding to the unit interpretation of the saccade length. Assuming I'm obtaining the “saccade length” by using the equation above (by column norm_pos_x & norm_pos_y), then I would get the result of the saccade length like the figure above.

Would like to know how to interpret the result into actual length ( e.g. mm or cm)? Many Thanks~

How can be the distortion be calculated if there is no undistorted Video? Using the gaze_point_3D seemd okay but comparing the plottet Datapoints it felt like there is an inversion of the Data an using both eyes, the Datapoints are quite seperated.

Hi@user-80316b. I'm not sure that I understand the question fully. Can you elaborate a bit more on exactly what you are trying to do?

Hello, I would like to ask something related to the calibration process: 1. It turns out that we have 2 ways for calibrating, and I just want to confirm if I did that in a correct way : (1) Screen Marker Calibration Choreography: keep the head still while moving eyes to track the markers, and (2) Single Marker Calibration Choreography: keep the eyes still on the marker while making spiral movements of the head. 2. In terms of accuracy, do you think which calibration approach is better? 3. When I finished the Single Marker Calibration Choreography, I removed the eye-tracker and turned off the Pupil Capture, but when I open them again, it is still able to detect my gaze, is it true that the calibration parameters from the previous process has been saved?

Hope to receive your support!

Hi! I'm using Pupil Core to study the glare of athletes during sports competitions. I am trying to find the angle that consists of the eye, the gaze point, and the lighting. Regarding the position of the illumination, the position coordinates (pixels) on the image were obtained by annotating the distortion-free image acquired by iMotios. The gaze position uses Gaze2dX and Gaze2dY acquired by iMotios. The angle is calculated assuming the eye position coordinates are at the center of the image (640,360). Is this calculation method correct? Thank you in advance!

Hey, I'm trying to extract all frames from the world video to jpg format using ffmpeg as described in this tutorial :https://github.com/pupil-labs/pupil-tutorials/blob/master/07_frame_extraction.ipynb. However, trying to account for variable frame rate using the -vsync 0 option gives me the following error:

Encoder did not produce proper pts, making some up. [mjpeg @ 0x7f982370c180] Invalid pts (41) <= last (41)trate=N/A speed=2.12x
Video encoding failed Conversion failed!

Using png as output format or using -vsync 2 (variable frame rate) works, but i'm not sure whether frames are extracted correctly. Could you tell me why this error is produced or maybe whether -vsync 2 is also a valid option for extracting frames? Thanks a lot!

Hi, I feel like distortion is less of an issue in this case. The reason one usually needs markers is that gaze is predicated in scene camera space but the latter can move relative to your AOI. You won't be able to map gaze to your AOI unless 1) you assume that there is no head movement relative to the AOI or 2) have some kind of external reference to calculate the relationship between scene camera and AOI over time.

I would also like to emphasize the difference between pupil and gaze data. The former is estimated in eye camera space and does not require a calibration. The latter requires a calibration and is estimated in scene camera space. You seem to have plotted pupil data if I am not mistaken.

The only way to be sure of this would be via surface tracking markers. Personally, I wouldn't be comfortable making the assumption of lower vs upper thirds of the video frame being associated with phone and world gaze, respectively, and would instead adopt manual annotation to discern whether or not participants gazed at the mobile device. Of course, this depends on how critical these insights are! 🙂

The manual annotations in Pupil Player do not have a duration. They are timestamped with the timestamp of the displayed scene video frame. I would rather recommend using two separate events, enter and exit.

Thank you @nmt for your information, it is clear to me know. But when doing the calibration with the physical marker, I am not sure when (how long) should we stop the calibration process? Secondly, after calibration, there is a spiral displayed on the visualization as images below, what does it mean? Does it indicate the religion of confidence for gaze detection?

@nmt Thank you for your fast reply. I also think that the effectiveness of the angle of composition needs to be discussed. Therefore, I am currently studying the relationship between the installation position of the lighting equipment, the pupil contraction rate, and the subjective evaluation. Is it possible to convert eye_center * _3d_x, y, z (* eye0 or eye1) to 2D? I processed the distortion-free 2d image plane output by iMove Exporter in Pupil Player.

Thank you @nmt , then is it true that it is better to let this spiral cover the whole field of view (of the world camera)?

Unfortunately, it fell off my radar. I made a note and will look at it first thing tomorrow morning

Hi @papr (Pupil Labs) . I want to manually save and load 3D eye models, so I tried to dump self.detector into a file in the pye3d_plugin, but I got a RuntimeError: SynchronizedArray objects should only be shared between processes through inheritance. I also tried creating a method in detector_3d.py (the detector class in pye3d package) to access the 3d eye model, but I still get the same RuntimeError. It seems like an issue related to multiprocessing, but I cannot find anything related to multiprocessing in pye3d class (in file detector_3d.py) or pye3d_plugin.py. Could you please point out how I should solve this issue?

Hi. Sorry for the delay. Multiple people have been using the system. Here is the current state of things: - in Device Manager, the Pupil devices appear as a single device: "Unknown USB Devices (Device Descriptor Request Failed)", under "Universal Serial Bus controllers". - We have another, monocular Pupil headset, which when plugged in appears solely as two libusbK devices, and works fine with Pupil Capture. - The binocular headset works perfectly with Pupil Capture on another machine. This issue started after someone uninstalled the drivers for the binocular headset in Device Manager, and attempted to reinstall them by running Pupil Capture as administrator as recommended on the website. Prior to reinstalling the drivers, the binocular headset worked perfectly. Uninstallation/reinstallation as described on the site does not help.

I guess something about the device configuration isn't getting cleared out upon right-click->uninstall. Since the device is unknown, there's no checkbox to "delete the driver software". Is there a specific directory I should look in to find and remove such files? Thanks again.

I see. Thanks! I will study a little more about the modeling code and try to dump things out in other forms.

May I ask that is it possible for the users wearing eye-lens to wear the eye-trackers? Also in case of wearing eye-glasses?

Hi, another quick question in terms of the distance between the eye and fixation, is it correct if I apply the "eye_center0_3d_z" in "gaze_positions.csv" & average of ”gaze_point_3d_z" in “fixation_csv" to obtain the linear distance ?

Many Thanks

The depth component of gaze_point_3d is prone to inaccuracies, particularly at longer viewing distances. gaze_point_3d is defined as the intersection (or nearest intersection) of the left and right eyes visual axes. Predicting depth from these is inherently difficult. For a more accurate measurement of eye-to-fixation distance, I would recommend using the head pose tracker plugin: https://docs.pupil-labs.com/core/software/pupil-player/#head-pose-tracking

I fixed the typo in the docs and but was not able to reproduce the issue. Hence, I created this debug plugin https://gist.github.com/papr/98636456fd4b4d0621835ac0d3cffc77 Use it instead of the official plugin. It will log an error message when it runs into the issue above instead of crashing. Please share the message with us should you run into the error again!

Here is how you can install the plugin https://docs.pupil-labs.com/developer/core/plugin-api/#adding-a-plugin

If your exported data has no head pose tracker results, then that option is off the cards. How accurate does your eye-to-fixation point distance need to be? I wouldn't rely on it for anything important, but rather use it as a rough estimate

Note that there is no way to check the quality of the distance measure without access to head pose data

You can work directly with the gaze_normals to calculate saccade amplitude in degrees

Hello. I want to export some data again because I changed something. During the eport I get an error message: "surface_tracker.surface_tracker_offline:Surface Gaze mapping not finished. No data will be exported". Any idea how to fix this? I tried recalculating the Gaze Mapper, but to no success.

Hi, I would like some advice on entering areas of interest into a video within pupil player. Unfortunately using surface tracking is not possible in our case (or we are not able to place markers in the scene). The issue is that we are doing interviews and need to compare the level of eye contact within the interview. Is there any way to manually insert the areas of interest into the video? Unfortunately I have not found a manual anywhere that addresses this issue. Thanks

Thanks for sharing the figure – this approach is more useful for defining the size of stimuli, e.g. that are presented on a screen, in terms of visual angle. If you are interested in saccades, Pupil Core already provides theta and phi for each eye in the pupil_positions.csv export! Their unit is radians. Note that theta and phi are relative to the eye camera coordinate system.

The recording is 28 minutes long. Is there an indicator for the process of the surface tracker mapping the gaze?

Hi, how do I calculate common gaze from gaze 0 2d and gaze 1?

gaze from individual eyes. So I have gaze information from gaze.2d.0. and gaze.2d.1. How do I calculate common gaze from gazes for individual eyes?

Its working brilliantly. Waht is the meaning of two numbers in norm_pos ?

And also, would like to know how to measure the saccade amplitude between "fixations" rather than "gaze", what I understand is that the parameters "theta"&"phi" is based on the gaze

Hi, what video are you trying to extract frames from exactly? The intermediate scene video or the exported one?

A simple way to check if -vsync 2 works as expected would be to use this Pupil Player plugin https://gist.github.com/papr/c123d1ef1009126248713f302cd9fac3 It renders the frame index into the exported world video. After transcoding you can check if all frames are there as expected.

Thank you. I will have a look at it tomorrow

I can recommend reading "Identifying fixations and saccades in eye-tracking protocols": https://doi.org/10.1145/355017.355028 The most basic filter suitable for detecting saccades is probably the I-VT (section 3.1). Section 4 of this tutorial: https://github.com/pupil-labs/pupil-tutorials/blob/master/05_visualize_gaze_velocity.ipynb shows how to calculate angular velocity from theta and phi. Prior that operation you might want to convert theta and phi from radians to degrees

just wanted to follow up to see if your colleagues had any advice for how to go about fixing this issue, thanks!

Hello. Is it possible to create different forms of Surfaces in the Surface Tracker then an rectangle? For example an U shape? There are only four points available for changing the form, is it possible to add more?

Helo All. Is it possible to set surface mapper to include a bit more surface beyond mark boundary? or to use surface markers and adjust the shape (keeping it rectangular) of area we check?

@&288503824266690561 Hello, I'm currently trying to build from source, but I am getting an ImportError for uvc. Is there a way to fix this issue?