Hello! I submitted an article about my project where I used Pupil Core to make some experiments, but they replied asking for a permission to use one of the pictures. I guess it can be the figure that I took from your technical aspects section (the one with the glasses and corresponding components)... I put the reference of the website, but it seems is not enough. Is it possible to have a written premission from you? Thank you!

Please contact [email removed] in this regard.

I will make sure that you will get a response within the next 2 hours.

We have received your email.

Hi @papr When processing the pupil_position data exported from pupil player I exclude data what has a Pupil_confidence or Model_confidence below 0.6 as your website says that data with confidence below 0.6 is unreliable. However, this has resulted in a big loss of data because for some of my trials the Model_confidence for one eye (or sometimes both eye) is consistently low (<0.5, often <0.4-0.3) even though the Pupil_confidence is high (<0.9). I know that a low Pupil_confidence means that the pupil tracking was poor and it is often possible to see and fix this in real time using pupil capture during the experiment, but what causes the Model_confidence to be low and how can I prevent this problem from happening? Many thanks!

Okay, so in my case given that the pupil confidence is high the issue is probably caused by the a lack of data at different viewing angles?

This pilot study does yes.

The subjects are asked to fixate on a fixed target under different viewing conditions with breaks at set intervals. Each trails (and thus recording) lasts about 2 mins but the whole experiment takes about 40 min depending on the number and length of breaks the subject choses to take.

Freezing the eye model makes it susceptible to slippage. But if your subject is keeping their head still during the trial, slippage probability is low

Cool! Thanks! How do I check the models fit in real time? I might be missing something and but I can't see the model confidence displayed in capture?

and what is the advantage to freezing the model once it has a good fit (i.e. I thought it only update if a slippage occurs)?

Thanks very much @papr.

could Pupil enable my father to use his eyes to control a Windows PC and/or Android device? Or is there something you folks who understand this tech recommend for someone who has paralyzed arms? I'm sorry if this is the wrong channel to be asking this.

Hey, I’m trying to use pupil core on a windows computer and I keep getting device has malfunctioned pop up

Never mind, it worked

@papr hello, I was wondering how we can limit the information we are trying to gather from pupil core to the dilation of both eyes, their azimuth, and their virgence (if that is calculated by the machine).

can I apply host-hoc calibration from pupil-player for next pupil-capture recording session?

ok thanks!

Hi, could I get your advice please, I need to record 2 world videos simultaneously with different recording frequencies. Then I want to match times of the videos and present them next to each other. Is there something that I can use for the time matching?

I use pupil capture 3.1.16 on a win10 system.

HI all, I'm trying to setup Pupillabs 3.x with my VR application using LSL. I'm well aware that Pupil Capture needs to be calibrated before any data will be put out on the streams. However, I'm not interested in the gaze data - only Pupil diameter, which as far as I know doesn't need depth gaze calibration to work. Are there any plans to allow the LSL plugin to output pupil diameter data without calibration in the future?

I know I can include a calibration procedure in my VR application but it is already quite taxing to begin with so I'd like to avoid this.

Thank you for your prompt response papr! Right now I've rolled back to 1.23 but I feel like I'm missing out on new juicy updates. I'll see what I can do with implementing it myself - Thank you!

Hello, pupillabs does not save any annotations I made. The annotation.pldata does not safe or change even though I the annotations pop up when I look trough the video. Any ideas?

Thank you for your rapid answer. Pupillabs crashed and therefore the annotation file did not update. That was the problem...

hi, does the output data means normalized data?

@papr

Please what do these lines mean? Only the eye 0 has the red circular line

The calibration doesn't cover the whole screen

I need the latter. would you know how I can use PL world camera with any other recording software? I'll record the picoflexx video with the pupil capture.

hi, little update: somehow I got 2 sessions of pupil labs capture running on my laptop. one recording world view with your high speed cam and one with eye-tracking and picoflexx plugin

so I can use the timestamps to map both videos. matching field of view.... maybe adobe will sort that for me.

Did you perform a screen-based calibration, or did you use the physical marker (in the right of your picture)? If you did a screen-based calibration and the physical marker was also visible, this will throw off your calibration.

Hey, I'm just wondering if the LSL plugin has changed substantially. I went to use the most recent version today (2.1 I think?) with the newest version of Pupil Capture and the newest version of Lab Recorder and found that it is not recording any pupil position data, but the outlet still is open. It passes all of the metadata about the outlet, but there are no time stamps or timeseries data

Oh, I don't think I had to do that before

Even if I just want pupil location data - not gaze?

Yes, see this message for reference https://discord.com/channels/285728493612957698/285728493612957698/829283882300997674

Ok thanks

@user-ecbbea @user-af5385 I will note this down, though. Maybe we can create a pupil-data only LSL plugin since there seems to be demand for it.

yeah, I'm confused if I was just dreaming of it, but i'm almost positive I was able to get pupil position data out before. was this changed?

I would super appreciate it - in the meantime I'll try to hack together a pupil only version

ah ok. yeah maybe either a separate version - or just split the gaze and pupil data into two streams. that way you can access both streams (or just one) and still be synchronized to LSL's clock

ah I see. seems like a compromise could be to push pupil data as normal and fill in nans for gaze data if a good calibration is not available. but maybe I'm thinking too simply

it just wasn't clear I had to calibration before hand

yeah no worries - I know you're probably busy. in the mean time - do you have any quick tips for generating the dummy gaze values? can I just hardcore them into the original script?

Until two days ago, the glasses were well executed, but now the glasses themselves are not working. The two parts that recognize iris are good, but the camera on top itself is not recognized. I just can't figure out why. Please help me. ( ps. I am using window )

Is there a possibility to find out how long someone looked into someone else's eyes when doing the analysis? It's a bit complicated to patch a marker at each side of the eye 😅

And how high is the precision of measurement of a pupil labs core? - so how big is the aberration between the actual gaze and the measured gaze point, normally given in degrees?

Hi @papr , I have one query regarding the data output from PL software. 'circle_3d_normal's and 'gaze_normal's. They both are normals originating from the sphere center or eyeball center. 'circle_3d_normal' originates at sphere center and passes through pupil center in 3D and refers to the optical axis whereas 'gaze_normal's refer to visual axis but this also originates from sphere center. Is this correct?

Is there any way to get the matrix that converts eye coordinate space (ECS) to scene coordinate space (SCS) from the export files only? Is this trasnformation stored anywhere or do I need to do with calibration again?

Are you referring to the calibration points or the matrix?

I only see the calibration points in these files

It would probably be easiest to write a Capture plugin that loads the calibration data from a given recording, initializes the gazer and writes the captured calibration result to disk.

Actually, you do not even have to write a plugin. This works with an external script and the network API alone.

Thank you @papr. I will have a look at it.

I want to know what information the output data(like gaze position.csv) can deliver after you use pupil capture to record.

Hi, I am using Pupil Mobile + Pupil Player 3.2 . I found at the one timestamp point, Player crashes and found world_lookup pts reset at that point.

error message is like this. Any help?

2021-04-12 09:09:22,727 - player - [ERROR] launchables.player: Process Player crashed with trace: Traceback (most recent call last): File "video_capture/file_backend.py", line 535, in seek_to_frame File "video_capture/file_backend.py", line 185, in seek File "av/stream.pyx", line 146, in av.stream.Stream.seek File "av/container/core.pyx", line 166, in av.container.core.ContainerProxy.seek File "av/utils.pyx", line 105, in av.utils.err_check av.AVError: [Errno 22] Invalid argument The above exception was the direct cause of the following exception: Traceback (most recent call last): File "launchables/player.py", line 690, in player File "video_overlay/plugins/eye_overlay.py", line 57, in recent_events File "video_overlay/workers/overlay_renderer.py", line 51, in draw_on_frame File "video_overlay/workers/frame_fetcher.py", line 34, in closest_frame_to_ts File "video_overlay/workers/frame_fetcher.py", line 43, in frame_for_idx File "video_capture/file_backend.py", line 537, in seek_to_frame video_capture.file_backend.FileSeekError 2021-04-12 09:09:22,727 - player - [INFO] launchables.player: Process shutting down.

Please I have questions

The green lines must be within the ROI right ? And should be smaller? I am still trying to look for the range of values that are ideal for accuracy and precision.

After calibration must the the field cover all the stimulus areas ? I am going to send a picture to understand this last question

Must this cover the whole range of the visual stimulus ? The participant will not move the head during the tracking

Thanks. Is this possible when the visual stimulus is very large like 95 inches ? The visual stimulus is placed on the wall. Is single marker calibration good for such calibration?

Hi! I have the following issue with the surface mapper and I would like your advice. I have an A4 paper with 4 april tags in the corners. I have defined a surface on pupil player, with all 4. However, it does not detect the surface when only one is within view. I thought the surface works even with a subset of the apriltags visible. Is there some kind of rule of thumb? 2/4 minimum or so?

edit: to clarify, when there 2-3 markers visible the surface is detected, and has the annotation <surface name> 3/4

Thanks

Can I ask how to get the gaze coordinates of the calibration points? Is that possible? I want to draw a bounding box as a limit for my experiment.

That is possible. When stopped, the calibration choreography starts the selected gazer plugin using a notify.start_plugin message. It includes the pupil and reference data necessary to calibrate.

Can pupil lab core audio-recorded? Like participants able to indicate when they actually see the target? This will help us know what is correct and incorrect

Do we have filter for setting threshold for what should be considered fixations or saccades?

Ah, my apologies, I was assuming you wanted to do that in real time. You wanted to do that post-hoc only?

I am not 100% sure on that.

Where do you want to draw the calibration area?

And do you use post-hoc calibration? Or the recorded gaze data?

Regarding the real time do I have to enable a plugin?

Check out this quick notebook I made https://nbviewer.jupyter.org/gist/papr/ad50c1146d297deef9a1738a4731eb45

Thanks, that's not the issue. the individual marker is detected fine, but when it detect 1 out of 4 only (the others are out of view) it does not 'show' the surface.

i see!

in another setup we tried (successefuly i think) to define a surface from a single marker. this seems to work fine>

no, it also worked on the pupil player

I can pm you some examples

I was just able to reproduce it. My apologies.

So, it seems like one-marker-surfaces need to be setup as such. Surfaces defined with more than one marker require at least 2 surfaces.

any chance we can tweak this in the plugin?

it would make detections somewhat smoother

I am not exactly sure why and how the software differentiates these cases. We will have to look into that.

ok

let me know if you find out anything 😉

Thanks again! Because pupil lab core has a binocular recording, does it mean we will have two data? One for the right eye and the other for the left eye? If that is a yes, how can we decide which data to choose? Right eye or average the two data ?

For the offline fixation detection what does this statement mean? " Instead of creating two consecutive fixations of length 300 ms it creates a single fixation with length 600 ms. Fixations do not overlap"

Is the minimum duration threshold of 100-200 ms proposed by Salvucci for online fixation detection?

https://www.frontiersin.org/articles/10.3389/fnhum.2020.589502/full#h3

In this paper, the authors used screen marker calibration for experiment that doesn't involve the use of the computer screen. I want to be sure if is possible to use the 5 point screen marker calibration for experiment where the visual stimulus is placed on the wall ?

Hello! I am using the pupil core eye tracker and since my last recording the world camera does not work anymore. When i start pupil capture the screen for the world camera stays grey.. I tried to reinstall the cam driver but it didnt help. I guess the problem could be a loose connector which connects the world camera to the usb hub. The other 2 cams for the eyes work fine. I am grateful for every tip!

@user-7daa32 that depends on what outcomes you are interested in. If you simply want to know where people are looking, the gaze point is estimated from both eyes automatically (binocular). If you want to examine the specific gaze direction for each eye individually, you can also do that by looking at 'gaze_normal*_x,y,z': https://docs.pupil-labs.com/core/software/pupil-player/#raw-data-exporter

Please contact [email removed] in this regard.

@user-7daa32 here are a few example scenarios: 1) Your stimulus is 2x2 meters in size. The participant stands 2 meters away from the stimulus. In order to look at the whole stimulus, the participant will 'point' their head towards where they are looking. Therefore, the degree of eye rotation in the head won't be big. 2) The same conditions. However, you instruct the participant to keep their head still. Therefore, in order to look at the whole stimulus, they will need bigger eye rotations in the head. The calibration procedure you choose should replicate the eye rotations. So if your experiment aligns with scenario 1 (i.e. the participants are free to move their head) a screen-based calibration will probably be sufficient. If you experiment aligns with scenario 2 (i.e. the participants head must be still) you might need to use a single marker in order to calibrate at the bigger eye rotations.

To be sure, you should run a pilot study. You should try out both calibration methods. Then confirm which one was optimal for you experimental setup. Test them out by instructing the participant to look at a specific region and check the gaze point is in the expected place.

zigzag — Today at 11:14 AM Hi, a couple of questions I have a pupil labs 200 hz core eye camera.

I would like it to be positioned pointing straight at the eye and quite close (60 mm away) or so but I am having some issues getting a good image - perhaps the software is not expecting a head on view of the eye? I had a brief look but I don't really understand the plug-ins and whether or not an algorithm optimised for a head on view of the eye exists

My user will have glasses on so perhaps the glasses are causing issues?

Perhaps the fixed focus distance of the 200hz camera is too far away - what is the design focal distance ? I know it's not recommended, but could I adjust the lens focal position (or change the lens) I assume it is a standard m12 mount lens , and the glue fixing the lens could be scraped off ? (I accept all risk of destroying the lens)

What is the optimum situation for maximum contrast (i.e. dark pupil and white iris) - to have a high level of uniform illumination on the eye to increase the contrast ? (while keeping eye safety)

Does the software modulate the LEDs at all for glint detection (i.e. to distinguish between the glints from the two LEDs ?)

I assume the LEDs are at 850 nm, but I can still see a red glow from the LEDs which surprises me (perhaps the lower wavelength tail of the LED?)

I assume the pupil labs camera has an IR pass + visible block filter in it , and hence any replacement m12 lenses would need this too?

Thanks!

Hello! Is it possible to run Fixation Detector in console (not via interface of Pupil Player)? If yes, please tell how (I cannot find scripts in the pupil repo)

A single press of a hotkey will only generate one annotation. Are you sure the hotkey wasn't held down or pressed twice? Incidentally, I would avoid using 'E' as a hotkey, as 'e' is actually used to trigger a data export in Pupil Player

When using a single marker, yes you can move the head, e.g. in a spiral pattern. Alternatively, you could move the marker around whilst keeping the head still. The most important thing is to cover a similar area of the visual field that you will record in your experiment. The 5-point screen-based calibration can only cover a relatively small area that encompasses the computer screen. With the single marker, you can cover a much larger range of the visual field as you can move it or your head as much as needed.

it looks like my camera intrinsics calibration may have been off during data recording. The displayed image looked correct (all parallel lines parallel) but when i go to undistort the frames in post processing, the undistortion fails. If i load the default intrinsics I am able to somewhat undistort the image while if I load the calibrated intrinscs I just get a black image when I display the undistorted image. If I recalibrate, is there a way i can still use the gaze projections? or will these now be off for all data collected? (I am only interested in 2d gaze data for the moment)

2d gaze data is not affected by the intrinsics as it is mapped into the distorted camera image.

Hey, which camera model have you chosen for the intrinsics estimation? Radial or fish eye? Please use the latter for the 1080p resolution.

Hi, anyone has the wiring diagram of the core? I need to change the connector of the right camera.

Please contact info@pupil-labs.com in this regard.

Hi all, we're using the core to run an experiment in the light and in complete darkness, but in darkness we are struggling to get a reasonable confidence level, and we are missing a lot of data. Our hunch is that the pupil is probably too large, but I thought it would be a good idea to see if there are recommended settings for collecting data in darkness in case we are doing something silly! In the light we have nice data, so the only difference is in darkness and we're a bit stumped. Any advice would be very much appreciated, thanks.

Please see this issue for reference https://github.com/pupil-labs/pupil/issues/1370

Hi is there a link to latest update for Pupil Core? Thanks Gary.

Thanks Papr, hope you are all well.

Yes just trying to get my brain working again after the long gap 😂

Sorry to bother you! But how can I get the concrete bounding boxes on the images by the gaze data, pupil data and other data?

@user-a9e72d what are you looking to achieve? Are you looking to map gaze data to surface/image in the real world? If so, you might want to see: https://docs.pupil-labs.com/core/software/pupil-capture/#surface-tracking

@wrp Sorry to bother you again. After I got the gaze position and pupil position. I want to use the data to get the area which has been looked at by users. (The position can be mapped to the screen?)

I did not find related document!🥲

Thank you for your kindly help

Hi all,

I was wandering if someone already has troubles with the sampling rate of the pupil glasses ? I use LSL to record pupil and EEG data at the same time but the sampling rate of my pupil core is changing constantly (from 180 to 260Hz !). While the EEG sampling rate stay constant (around 512Hz). Did someone already has this kind of problem ?

Good question, I'm not sure but I use an old version of pupil Capture (version 2.1 or 2.2).

LSL calculates the sampling rate based on the data timestamps. In the first plugin version, these were not measured at frame exposure but after performing pupil detection. The variable duration between frame exposure and completed computation caused the uneven timestamp distribution, and therefore the variable "effective" sampling rate. The latest LSL plugin attempts to use the frame exposure timestamp instead, yielding a much stabler sampling rate calculation.

The Capture version is secondary here. The plugin version is more important. But given that Capture version, I am assuming that you are using the old plugin as well.

Ok, I understand, thank you for the explanation ! Where can I download the latest version of the plugin ?

Yes, my participants perform calibration before starting the recording. But what is a valid calibration ? There is a threshold that indicate if the calibration is good or bad ?

Thanks for the link

Ah yes, I see thank you. I have another question, regarding the "Nan". In the time series data I sometime have "Nan", does it meens that the glasses cannot capture the pupil at this particular moment ? What is the impact of the calibration on the quantity of Nan in the data ?

Ok this is the reason why I have Nan values in the diameter values. Probably because the device cannot detect the pupil.

I have a quick question about dispersion filters to be fed into the pupil detection algorithm. I have seen a lot of papers that talked about the duration threshold which depends on the research questions, nature of task, stimulus design, field, literature or researcher's discretion but I have not found explanation for the dispersion threshold. I read that degree of visual angle depends on the tracker resolution. But I don't know which ranges of spatial bound should be taken as fixations or saccades especially for pupil lab core.

For example, in chemistry education, most visual stimuli have multiple representations required to solve a task and so the duration threshold usually set at 100 Ms.

Does detect eye0 and detect eye1 need to be checked, while recording? Or can they just be used to position the eye cameras correctly?

No eye recording seem to have been made, when they are unchecked.

When would the next release be? Are there any releases that I would be able to use in the meantime? the laptop that we are using is maxed out when they are on.

Thank you for your help! 🙂

Hi @papr,

What would be the minimum specs for running the software?

We have an i7-7500U with 8GB ram, but we are having issues with this

The coarse detection is part of the 2d detector. It is disabled at the lowest eye camera resolution (eye window -> video source menu -> resolution). If that does not help, please share an example recording with [email removed] such that we can have a look.

is there any update on this by any chance? I turned out that participants look at the stimulus from more close so some tags are out of the field of view; there is at least one though and it would be great if we could still use it to detect the interaction with the surface

Hello! I have a question about the frame rate. I exported the fixation data, and I believe the frames used their are for frames at 30hz, is that correct? Could you briefly explain how it goes from the 120Hz sampling rate of the world camera and 200Hz sampling rate of the eye cameras, to fixation frame rate of 30Hz? And, is that modifiable? Thank you so much!

I'm trying to do some analysis on the video image statistics at each fixation. Since I have the start frame and end frame of the fixations, I can just pull the frame from a simple video split function. However, I need to know what is the sampling rate of the frames indicated in the fixation file. Does that make sense?

great! I will look into what you have sent me!

Okay, so if I understand correctly if I extract frame index 4 through the process outlined here: https://github.com/pupil-labs/pupil-tutorials/blob/master/07_frame_extraction.ipynb, and then get the fixation xy coordinates at frame 4 in the fixation file, then I should have a good corresponding frame to fixation right?

Yeah, it is true a fixation can last many frames, but let's say that for the sake of argument, a fixation happens in frame 4 and only in frame 4. Then, I can use the method described in that tutorial to pull that frame from the video correct? Thank you so much for helping me figure this out 🙂

Hello

Thank again for the help.

1. I have another questions about calibration. Can I just have like 4 to 6 calibration markers placed on the wall instead of moving a single marker around during calibration?

2. How big can a surface (AOI) be to have actual or correct fixations within and not in whitespaces or not interested areas? Participants should claim they are looking at A for example but the fixations seen in player are on B or in between A and B.

3. If we have the Alphabets as the stimulus for example and participants are to search for a particular letter as commanded, how big do you think each letter could be? How big in respect to the tracker resolution

Hi everyone, I am currently working with the Surface tracker plugin to detect April. And to detect AprilTags it is mentioned that you need a grayscale image of type numpy.uint8. I took a look at the code of the plugin and it appears to receive a "frame" which must be of numpy.uint8 type. So I'm wondering where this "frame" is coming from and what is it's real type.

Please could someone help me with that ?

Thanks for answering quickly. Sorry if I was unclear, I am attempting to run the marker detection. I saw it is using this code to detect April tags : https://github.com/pupil-labs/apriltags/blob/master/src/pupil_apriltags/bindings.py line 400 for the detection part

And I saw that finally the image which is needed comes from here : https://github.com/pupil-labs/pupil/blob/master/pupil_src/shared_modules/surface_tracker/surface_tracker_online.py but I don't know where this frame come from

So I'm trying to get the image from the glasses and directly give it to the detector.

Capture uses https://github.com/pupil-labs/pyuvc to access the cameras here https://github.com/pupil-labs/pupil/blob/master/pupil_src/shared_modules/video_capture/uvc_backend.py#L459

Yes that would be better

Thanks ! I'll take a look at it

This happens a lot lately to my recordings, that for some time during my experitment, cameras don't record anything and they suddenly need to reinit. How can I prevent this?

Hi, I'm new to pupil. Is there any webcam support?

I am the only one who has a problem with the program Pupil Player to not be responding - even though CPU is at 13.1 only? All programs are closed and this happens always. I managed to export data once. The video is only 2.5 minutes long ( a test video ) but even this is too much? Help please 😟

I use Pupil Player v3.2.15

funnily, I just opened v3.1.16 and there it works 🤔

Looks like a physical disconnect of the cable since all three cams disconnected at the same time.

Hi everyone, is logitech c270 compatible with Pupil Capture v3.2.20?

It's the first time I try it, it doesn't see the webcam at all

@papr Win10

Yes I am following those instructions, as soon as I solve an error with pyhton in the last step I will have my answer

I'm running the compiled version named "Pupil Capture v3.2.20"

After following all the steps in the guide for custom setup world and eye0 cameras can't seem to recognize the c270, it says 3 times"unknown"

I'm sure I've installed the drivers manually using Zadig

Ok I figured out what I did wrong:

When using Zadig, once found the correct composite devices I did not set manually the driver to install to the "libusbK (v3.0.7.0)" (my bad), to everyone having issues with drivers resulting in 3 "unknown" cameras this could be a mistake.

Now I have another issue, I have 2 c270 to use, is this an issue? (when the 2 cameras are basically the same?) Because both World and Eye0 use the same one on my eye and the one looking forward is not used

That would not be a problem, but in my case both world and eye0 or eye0 and eye1 open only 1 of them. Basically they take only on of the 2 I have. I have installed the correct drivers on both (I've checked to be sure)

I understand but I've tried all. Anyway the front camera is used only for Eye Tracker feature right? I am more interested in a pupilometer

Please I have trouble with the gaze not hitting the AOI being looked upon by the viewer. For example, let say we alphabets A to Z. Each letter is an AOI. The AOI is made of a letter surrounded by a 5” 5" squared black color line and again by a 1.2" 1.2" squared whitespace. I guess you get it!

All placed side by side in order of the Alphabets arrangement. By so doing the whitespace between two 5"5" will be 2.4inches.

The question I wanna ask is this.. How big or small can the size of an AOI be for it to be hit and not another or whitespace? What size can account for the tracker error?

Capture turn off immediately after opening... I don't know what I wanna do please

Correct, you do not need a scene camera if you want to estimate pupil size.

You can calculate the stimulus size s using your calibration accuracy E and the target distance d.

What ever you want to identify reliably. 🙂 In your case target stimulus = aoi = letter

Sorry d is the size of the whole visual stimulus like the size of the plasma screen?

No, d is the distance from the subject to the aoi

The further the subject sits away from the stimulus, the bigger it needs to be.

I am asking a lot of questions.. Is the angle twice E? Or the 2 in the formulae has accounted for that ?

Thanks. I am currently considering that I might have the stimulus on the screen instead of on the wall. Please the AOI size we are talking about from that calculation, is it the size of of each feature on the stimulus or for the surface to be created using the Apriltags on player? I don't like creating surfaces on Capture.

There is the actual physical stimulus size and there is the AOI size place on the stimulus using the Apriltags

Thank you. Is is the size of the the tangible word on the screen or wall before data collection or the size of the word in capture or player created by Apriltags? I am talking about creating a surface.

Are you referring to the size of surface created ?

Or the size of the individual representations that form part of the stimulus?

Hello, could anybody tell me what kind of connector is used to connect the eye camera to the USB cable coming out of the Pupil Core frame?

You might be referring to the JST connector posted in this message: https://discord.com/channels/285728493612957698/285728493612957698/701705600979042336

Thanks. I guess the angular error is already known based on the tracker used right ?

I refer to the whole set up as the stimulus. And I refer to a relevant feature of the stimulus where surface will be created as the AOI. For example, whole A-Z is the stimulus. A is AOl if surface will eventually be created on A

So is the "your stimulus" the whole of the interested area?

After the calibration and validation, I will change the size based on the error ?

What if the size is fixed and can't be changed?

Maybe if fixed, then I Will have to vary the d right ?

Would the calibration remain valid ?

Another question I came across, I have noticed that in the pupil and gaze exports, some frames are occasionally skipped (e.g. I have an instance where I jump from world index 5460 to 5462, and a similar occurrence happens ~10 times in my 11 minute sample collection). The world frames are correctly captured (e.g. frame 5461 exists in my frames export, but no pupil data is associated with that frame) Is this expected behavior or could something be going wrong with my system?

Pupil data is associated by time. Can you check the exported world_timestamps.csv for the frames 5460–5462 and check how much time lies between them?

Hi, I have issues installing dependencies Could not install packages due to an EnvironmentError: [WinError 5] Access is denied: 'c:\program files\python36\lib\site-packages\pip-18.1.dist-info\entry_points.txt' Consider using the --user option or check the permissions. Is there a solution?

I want to get data from pupil core for my app

gaze data

@papr thanks

📣 Friendly reminder to Pupil Core users & developers about some useful Github repositories:

• Pupil Community: This repository contains community contributed projects, forks, plugins, scripts using Pupil Core. If you have made anything and would like it to be added to this repo, please make a pull request directly to this repo. 🙏 https://github.com/pupil-labs/pupil-community
• Pupil Tutorials: This repository contains a series of Jupyter notebooks that demonstrate how to visualize and analyze Pupil Labs eye tracking data. https://github.com/pupil-labs/pupil-tutorials/
• Pupil Helpers: This repository contains utility scripts and concise code examples that demonstrate how to communicate and develop your own tools with Pupil. https://github.com/pupil-labs/pupil-helpers

Hey, I'm seeing a pretty big offset of ~60 seconds between timestamps made on two different PCs on the same network, using LSL. One PC is running a custom lsl outlet which uses liblsl.local_clock() and the other has PupilCapture with the lsl plugin running. Is this a know issue? I don't believe any interfering pupil capture plugins are running. Data is recorded with LabRecorder and parsed with Matlab's open_xdf.

Also, what reference do you use to know that they are out of sync?

Ok this is driving me insane, no matter how I position the left eye camera I get pupil dropout for no discernable reason.. I've played with the intensity / min / max settings and it doesn't seem to make a difference. What am I doing wrong?

I'll give that a try - I've set them both to auto, which I thought would fix that

I think that may be the case, as reducing the exposure seems to make it worse

apologies for the second message, I was just wondering if you may have any further tips for debugging this issue. Thank you so much for your help.

I took a picture, and both of the IR LEDs seem to be working... ignore the mocap marker

I've been playing around with this for hours, using different computers, trying all of the angles I can and I keep getting the same problem... I'm not even really going into the extremes of my eyes. It looks like the IREDs are working, but I'm not 100% sure..