HI there, Im having trouble with the surface recording, the surface markers are set and saved, and the arrow is the right way, but there is nothing in the exported gaze_positions_on_Surface_[screen name] file. Any suggestions?

Thank you! 😊

Hi, I'm currently looking at someone else's data. This data has the strange effect that when surface data is exported, the x and y coordinates on the surface are perfectly correlated (they're not the same, but almost). This can also be seen in the surface heatmap that can be displayed in in the Pupil Player - It is a straight line from the middle of the surface to the top right corner of the surface, with most of it located in the top right corner. I have never seen such an effect before with my own data. The surface seems OK, and the gaze coordinates as well. But when compared, the output surface coordinates do not match the expected result at all. Is there something obvious that I have overlooked that can result in such surface data?

Hello! Would like some guidance here - this may be a "noob" question we are new to eye tracking. We have a Pupil Labs Core and want to run an experiment with users looking at a screen (iPad or Mac) and having the ability to use the eye tracker to move the mouse cursor on the screen (mouse cursor follows the gaze) and click on any buttons (either by fixating their gaze on the button for a specified duration or by blinking some number of times). Is doing something like this straightforward, and are there any guides for setting this up? I saw this online (https://docs.pupil-labs.com/alpha-lab/gaze-contingency-assistive/) but it seems like it's for Neon, and we have a Core.

Hey guys, I am trying to load previously used surface definitions for other recordings. I have copied the surface definitions v1 file into the other recording folder. However, when I load the recording into Pupil Player, it takes a long time to load with a white-ish screen. I have left it running for one hour with no success. I note here (https://docs.pupil-labs.com/core/software/pupil-player/) that it says it may take a few seconds because the surface tracker plugin and assessing the whole recording. My recordings are about 5 minutes long and I have 25 predefined surfaces. Is this normal for it to take this long to load?

Hi there. We are running into a continued issue with our pupil labs. We are running Pupil Core eye trackers through pupil capture v3.5.1. The glasses are wired directly to a laptop. However, upon starting, the right eye camera won't connect. Error message says "no camera intrinsic available" (among other things).

This error is consistent across both of our pupil core trackers, and occurs for both our Windows and Mac laptops. Any help would be appreciated!

Hi @user-f43a29 , I tried this, it doesn't change anything. Disclosure: I'm the lead developer of Blickshift Analytics and I received the data from a customer. Our Software allows importing the Pupil Labs Surface Data into our Dynamic AOI Editor, and the Dynamic AOI Editor also allows transforming the gaze coordinates into the local coordinate system of the surfaces/AOIs. Surprisingly, if I do this, the result is correct, i.e. if I use the surf_positions together with the gaze_positions exported from Pupil Player, and do the transformation calculation on my side, I get the expected result, but it does not match the content of the gaze_positions_on_surface file.

From my side, I have solved the problem by telling our customer to use the Dynamic AOI Editor and import the surfaces. His deadline is today, so it won't make a difference if I can find the source of this problem, but I would still like to understand what is going on.

Hi @nmt , I have received permission from the customer to share the data. I will send a link via pm.

How do I do this? (sorry)

Hi @nmt , weirdly enough when I tried again today, the surface definitions were able to load in a few seconds! However, there are still a few recordings for reasons unknown, which are not able to load the surface definitions in a timely manner (left on a white screen and loading for a long time). Even though these recordings are not longer than the others.

What does it mean when Pupil Player says "surface gaze mapping not finished"? I get this message occasionally when trying to export fixation data

I think it's also possible that the marker caches were nearly finished building when you reopened the recordings, but I can't say for sure. That message is literal. It means the surface mapping has not finished. It's recommended to wait for it to finish prior to exporting the data.

@user-d407c1 Hi Miguel, I see that the ticket we had open has closed. Were you able to get that mp4 file working, or did it not work out? Best, Kieran

Hello! I use lab streaming layer to record pupillometry from my pupil core and align to triggers from an exteral device. However, I want to record from pupil capture and analyze in pupil player as well. Do you know if I can record the pupil core video with lab streaming layer so I can align my data with triggers?

Hi @user-fc393e ! Your ticket was close due to inactivity, as we did not receive a response for more than 10 days. We tried several times to communicate there, basically my colleague @user-cdcab0 found the issue and replied to you there on the 2024-04-10. I am recovering his response for you here:

Hi, @user-fc393e - the crash you're seeing appears to be the result of a single corrupt record in your gaze data. I have written a custom plugin for you which, when enabled, will apply a patch to the fixation detector. The patch simply tells the fixation detector to ignore corrupted gaze records.

• Close Pupil Player
• Save the attached file to pupil_player_settings/plugins/fixation_patcher_kieran.py
• Start Pupil Player and enable the "Fixation Detector Patcher" plugin
• Enable the regular fixation detector plugin

@user-d407c1 already mentioned our list of best practices, but I'd like to also call attention to the Record Everything section. If this recording had included calibration like the recommendation, you'd be able to regenerate the gaze data post-hoc, which would have fixed the corrupt data.

Quick question: Is this fisheye effect in the Pupil Core world view image normal for a 1920x1080 resolution?

Hi @user-94183b, yes, the fish-eye distortion occurs when you select 1920x1080 resolution for the world camera. You can learn more about the camera intrinsics estimation here. You can also learn how to undistort images of the fish-eye lens by following this tutorial.

Hi @nmt , I understand what you are saying (this is always an issue with this customer, I have talked to him repeatedly about this during prior projects - to no avail it seems ;-). But this is not the issue I am referring to. I am talking strictly about frames where the markers are recognized, the surface is tracked correctly and gaze data exists. Otherwise our software wouldn't be able to compute the correct surface coordinates, I only used data imported from Pupil Core for this. I have attached a screenshot. On the left is the video frame. You can see that the gaze point embedded int the video matches the gaze point overlaid by Blickshift Analytics exactly (blue circle). In the middle, the gaze data is mapped onto a screenshot of the surface (ignore the embedded gaze data, that's an artifact of the screenshot). The blue circle is the surface-local gaze point computed by BSA, using gaze data + surface data. As you can see, it matches the location on the video. The red circle is the surface-local gaze point exported by Pupil Core, for the same timestamp.

hello i wanted to know where can i find the video i just recorded with my pupil to drag it in the pupil player

As you can see, it matches neither the point we compute, nor the video data. This is what I cannot explain. The red line in the line graph is the y-coordinate of the surface coordinate output by pupil labs. As you can see, it is almost constant, very close to 1. The light green one is the same coordinate computed by BSA. If it was a problem with the base data, how would I be able to compute this coordinate?

yes it worked thanks a lot

Hi @user-f43a29 , I have just tested it, and this indeed works. Thanks for the help. I would still have liked to understand what is going on (especially since the customer has about 40 data sets where the surfaces result in these incorrect exports), but I'll know what to do, if this happens again.

Hi, @user-fc393e - you're quite welcome! I still have a copy of your recording and the export works fine for me. Are you on the latest version of Pupil Player? Are there any meaningful messages in the log file?

Hi @user-311764 ! On the pupil_positions.csv you can find a column named method which tells you what detector was used, on each row.

Onto gaze, you can select which one is used for gaze estimation (see choose the right gaze mapping pipeline). Note that if you followed our best practices and recorded the calibration, you would be able to run it the gaze mapping post-hoc and choose a different pipeline.

hello i need to do a presentation for students and i wanted to know if there is some games that are playable with an eye tracker.

ok thanks you

is it possible to control the mouse with the pupil core or pupil invisible ? if so, how to do it please

Hi, I am using the Pupil-Invisible and the Pupil-Core for my study. Unfortunately, there were problems with the recordings with the Core glasses, which I only found out when I analysed the recordings. The situation is as follows: The settings and calibration worked and the fixations can be seen at the beginning, but as soon as the test person starts riding his bike, the pupils are barely visible due to the sun and there are no more fixations on the images. How can I avoid overlighting? What exactly do I have to set? I made the settings and the recordings on my laptop. is there also a mobile phone app for the Core that I can use? With the Pupil-Invisible and the matching app, the recordings work perfectly under the same conditions. I need to use two pairs of glasses as I have to take two recordings at the same time. Is there any way I can edit the recordings that have already been taken so that I get the fixations? I need the recordings for my dissertation.

This is the error I'm getting

I went down a similar path looking for games that are playable by mouse and then controlling the mouse via gaze. Here's a little "shooter" horde game I made and played while we were at a conference a couple of weeks ago. It's played using a Neon and the real-time screen gaze package

Also with Neon and the real-time-screen-gaze package, here's gaze-controlled Fruit Ninja

oh thanks, do i need to install some python modules to use it or I just need to run the codes ?

so i don't need to use the real-time screen gaze that you used for neon for the core ?

ok thanks i will test it and come back i have some issues

Hi! I'm trying out post-hoc calibration with manually identified referents to try to correct for some slippage we saw over our experiment. Is there a way to check that this is having a positive impact on fixation estimates? I can see fixation detection number before and after the post-hoc calibration, but I'm not sure if there's a better way to determine if the referent calibration is actually helping? Or to quantify how much it may be helping (e.g., more so for some participants than others)?

Hi! I would like to know if there is a simple solution to automatize the export of fixation data with fixed parameters (i.e. dipersion, min/max duration). Thanks in advance for your help!

Pretty sure I've seen your posts on reddit!

I post a lot on my project recently

Hi, does Pupil CORE works properly with Mac? I am getting calibration errors.

Please get back to me.

Hello all, I am trying to do a study using the surface tracker plugin. In the study we give the subjects 15 images with four markers at the corners of the image. Each image has to be analyzed on its own, but to make things easier we thought of using the same markers for each image since between two consecutive images there's a quiz with no marker so we use the surface_events file to know when an image is displayed or not. However, we've seen that the surface recognition is not always accurate as you can see in the screenshots from pupil player. Could it be because the size of the images are different or is it something else? I can see the camera is not properly on perpendicular to the monitor, could that be it? We have also turned down the aperture time since the monitor was too bright and the camera had difficulties finding the markers for the calibration

Hi, @user-4c48eb - some notes:

1. Yes, the surface mapping is being warped because you've anchored the tags to the corners of the images, and those images are different sizes. When you define a surface, that definition consists of the relative sizes and positions of the markers with each other. Here, you're changing the relative positions of the markers with each other, and that needs to stay static. Alternatively, you could either anchor the tags to the corner of the screen (instead of the individual images) or pad your smaller images so that they are all the same size.
2. The angle of the camera relative to the surface is fine. This only becomes a problem when the angle is so steep that the markers cannot be detected, and you're not even close to that in these images.
3. I wouldn't rely on surface visibility events. If, for whatever reason, the surface isn't detected in the camera (participant looks elsewhere, bad lighting/contrast (even momentarily), too much motion blur, etc), then you'll have a surface event that doesn't correspond with the stimulus event it's supposed to. You might be fine, but I would certainly be careful to review all the data before using this method.

A similar (but less error-prone) method which also doesn't require you to manually annotate and doesn't require separate surface definitions for each image would be to define a surface for the quizzes (using a different set of apriltag markers from the stimuli, but each quiz can use the same markers/surface as each other). That way, instead of looking for the disappearance of the stimulus followed by its next reappearance, you'd then be looking for the appearance of the quiz surface followed by the appearance of the stimulus surface.

How can I use the surface on the quiz to see where the participant is looking on the image? You wouldn't. I only suggest the quiz-surface to provide more reliable surface event sequences in order to know which stimulus they are on. You will still need to define a surface on the stimuli themselves.

the dimensions of the image depend on the dimensions of the screen itself I'm not sure I understand this, but I'd like to re-iterate one of my suggestions. Determine the height of the tallest image you have and the width of the widest image you have. Then add a black frame to all of your images so that all of them are as tall as the tallest and as wide as the widest. Now all of your images are the exact same size and you can display your AprilTag markers just as you are now (anchored to the corners of the images) without having to worry about surface warp.

I think the best option would be to create different surface definitions for each image This certainly is a solution that would work, but I assumed you had a reason for avoiding it - like if there are many stimuli, defining a separate surface for each one could be very tedious.

do you think I could manually export the data from the frames where only one image is present, deleting the previous definition of the surface and re-adding each time? This would work too, but again, sounds very tedious to me.

Hi there, I am researcher using the pupil core glasses. I am building a unique set up for a research project, and need to find a way to connect the cameras to USB. I am hoping to buy a connector/cord where one side of the wire should be USB and the other should be the type of connector suitable for these cameras. I have tried buying a few different USB to JST (I think that's the connector to the camera?) cables, but the camera connector is the wrong size each time. I'm attaching photos of what I am talking about. Is it possible to get the specs for that camera side connector, or do you possibly sell a USB to camera connector solution? Thanks so much!

Hi @user-f43a29 Fancy seeing you round these parts 😂 Thanks so much! This is super helpful. Hope you are doing well!

Hey all, I'm with TeamOpenSmartGlasses & we're looking into eye+pupil tracking as a means of detecting the wearer's focus.

Does anyone have an assembled (or partially assembled) Pupil Core DIY for sale? Would like to save time building one, if possible.

P.S. Sorry if this is the wrong channel to ask!

Excuse me.

Is there a way to use Pupil Core in WINDOWS11?　Last year, the application did not work in WINDOWS 11.

Hi @user-5c56d0, thanks for getting in touch 😄. Pupil Core's software is working with Windows 11 and my system is also windows 11.

Please try installing the software again, and open a ticket in 🛟 troubleshooting if the application didn't work for you. We'll try to guide you through some debugging steps if that happens.

Hello, when trying to export a failed recording with Invisible using the Pupil Player, I get the following error message: "NumPy boolean array indexing assignment cannot assign 155640 input values to the 264126 output values where the mask is true" Is there anything I can fix to make it work?

Hi @user-0b4995 , the main error that caused the crash is above that one. It says that eye0_lookup.npy is missing. By "failed", you mean the software crashed during the recording?

Also, since this is respect to Invisible, I hope you don't mind that I will move these messages to the 🕶 invisible channel

I will check with the others. What data do you need exactly? You want to recover all data that was recorded until the point of the crash or you want to just recover a subset of data streams, like gaze+world video?

Also, have you uploaded this recording to Pupil Cloud? What is its status there?

Well I just copied the eye1_lookup.npy' and eye0_lookup.npy' from another recording and it seemes to work now!. Skipping through the files and the gaze looks good

correct! i will check the export but since we code this one manually and the gaze seems to make sense I suppose this should work in this case. thanks for your help!

thanks rob!

Some quick DIY Core questions:

1. For the exposed and developed color film negative or Edmund’s 5-mm-diameter filter part, is this the correct part?: https://www.edmundoptics.com/p/5mm-diameter-optical-cast-plastic-ir-longpass-filter/41541/

2. I read somewhere that the accuracy of pupil diameter tracking heavily relies on your eye cam's resolution. Is the 720p Microsoft HD-6000 good enough for this, or do I need to modify the housing to accept another Logitech C615 (which is 1080p)?

Thanks!

Hi there, I was downloading your installation files for core on Linux Ubuntu 22.04 LTE. Installed it with Gdebi package installer and try to run via command line. However, I run into an error.

/opt/pupil_capture/libpython3.6m.so.1.0(+0x160a84)[0x7efd67160a84] /opt/pupil_capture/libpython3.6m.so.1.0(PyEval_EvalCodeEx+0x3e)[0x7efd6716106e] /opt/pupil_capture/libpython3.6m.so.1.0(PyEval_EvalCode+0x1b)[0x7efd6716109b] /opt/pupil_capture/pupil_capture(+0x372c)[0x55d7d5b8c72c] /opt/pupil_capture/pupil_capture(+0x3b1f)[0x55d7d5b8cb1f] /lib/x86_64-linux-gnu/libc.so.6(+0x29d90)[0x7efd67600d90] /lib/x86_64-linux-gnu/libc.so.6(__libc_start_main+0x80)[0x7efd67600e40] /opt/pupil_capture/pupil_capture(+0x24fa)[0x55d7d5b8b4fa]

Unhandled SIGSEGV: A segmentation fault occurred. This probably occurred because a compiled module has a bug in it and is not properly wrapped with sig_on(), sig_off(). Python will now terminate.

Could you please advise if you have other install instructions? or how to overome that error?

many thanks in advance.