Hi, i s it possible to subscribe to the normalized fixation points on a surface directly using zmq filter such as "surfaces.surface_one..."? If not, are there any implementations available for reference of this filtering in c++? Thanks#

so, i want to subscribe to fixations on surface named a, it would be

surfaces.a.fixations_on_surfaces?

Hi, is the pupil core MRI compatible?

Hi! I’m a beginner to this and haven’t messed with any code. I’m just trying to use the Downloadable software pupil capture for macOS and I’m having trouble. When I switch to local USB it’s telling me that the cameras are in use or blocked. How do I fix this?

Hello, How I can connect and use my eyetracker with smartphone?

But there is this Pupil Remote app right? Is that for tablet only?

hi I made a code to extract gaze and pupil data from unity, however it for some reason only works when used in conjunction with old pupil capture software (1.0) and when using pupil capture 3.51 it keeps giving me the error message "notification without timestamp will not be saved". Does anyone have any idea how I can fix this?

We have a loose wire (the red on) on the eye-tracking camera of out pupil core. How do we go about getting it fixed?

Can anyone tell me what is meant by Outlier Threshold in the validation section of pupil player and why its default is 5 degrees?

Hi team, I'm with Pupil. I tried a few lines of code to extract the position of the eyes and the gaze direction. I received a 3D position (x,y,z). What I couldn’t find is the position of the origin. Can anyone help me with that? Thanks in advance 🙂

First I used Pupil Player, and then I used a script with "suscriber.suscribe('gaze.')".

Just to clarify: With Pupil Capture you record the recordings which you can then export with Pupil Player. If you subscribe, you connect to Capture' realtime data stream. This might not be suitable for your analysis

Do you need to know where the subject is looking at in real time? Or is it sufficient to know where they were looking at after the recording?

In this case, subscribing to gaze is very suitable.

Hi, My question is probably answered somewhere but I can't find it. Does pupil labs use a mix of bright and dark pupil methods?

No, Pupil Core only looks for the the dark pupil. Reflections are not used.

https://arxiv.org/abs/1405.0006

Thank you for your response! I was also wondering how much of the green dot with the yellow rim and the center red dot that reflects where the user is looking has to be outside of a set surface tracker for that data point to be classified as FALSE for gaze_onsurface?

Hi all, could anyone help me for launching the python source code of Pupil - eye tracking platform? I followed the two instructions 1) "> python -m pip install --upgrade pip wheel" 2) "> pip install -r requirements.txt". During the installation, I got one error with the setuptools>61.0. And when I launched the application "> python ./main", I got an error calling git. Thanks in advance

Hi! Please note that you need to clone the source code. It is not sufficient to download the zipped archive.

Which operating system and Python version are you using?

Hi, sorry I come back with the same problem with launching the python source code of Pupil core. I found the full description of the installation. But in the end, when I execute the command "> python main.py capture", I receive the error message "ValueError : Version Error"

Hi, I am learning to use Pupil Core for the first time and have a question about the pupil detection and calibration. I followed the Pupil Core Quick Start guide but during the pupil detection the circles around my eyes are blue not green. Any suggestions on how to resolve this? Also, after calibration, I had a thin green rectangle with orange figures in each corner and a red circle in the middle. Refer to image below. Does anyone know what this is?

Hi ✌️ I am trying to calculate head position angles via the Head Pose Tracker Plugin, while looking on a 2d screen. I put 3 april tags on my screen but the calculation of the 3d Modell uses just one of the april tags. I think this is because the other two april tags are not visible all the time but I dont see why this would be a problem. Is there a way to manually select the markers or is this no problem as I am just looking at a 2d window anyway? Thanks in advance 😃

Hi all, I want to add the pupil core application to an existing Unity Application. So I added a button in the Unity Application to launch the pupil core (DID IT). Now I want that the pupil core is running in the background with no graphical interface. In other words, is there a clean way to remove the UI from the pupil core? The goal of putting the pupil core in the background is to allow me to collect data. Thanks in advance 🙂

There is a --hide-ui flag that you can pass when you call the exe. This will create a hidden window.

Hi, how are you I came to Germany from Mongolia and wrote an e-mail to get acquaint with your company and products, but I can't get in touch. What shoud i do?

Hi all, I encountered difficulties running the pupil core from the source code. 1) When I export the data, I do not find any data on the gaze_positions, but the rest of the data is OK (purpil_positions, world_timestamp,...). 2) I do not receive the stream data from subscribe('gaze'). These two problems do not exist when I run the app from the installed program. Would anyone have any idea? Thank you in advance 🙂

Hi @user-80123a! The behaviour you describe is usually indicative of no calibration. Did you calibrate the eye tracker when running from source?

I forgot, I cannot use calibration because the Eye Tracking I use right now does not have a frontal camera

But I will use a new eye tracker soon, and I hope my code will be ready when the new eye tracker arrives

To get gaze data, it is necessary to calibrate. Check out the docs for future reference: https://docs.pupil-labs.com/core/#_4-calibration Indeed, a scene camera is also necessary 🙂

The installed software will use the last-run calibration. This won't be accurate in all likelihood, unless the headset went back onto the same wearer and into exactly the same position (highly unlikely). The calibration is wearer-specific. If the headset is put onto a new wearer, or removed and re-worn, we recommend re-calibrating. Read more about calibration best-practices here: https://docs.pupil-labs.com/core/best-practices/#avoiding-slippage

Hello, best productive days to everybody 🙂

I'm going through the documentation for a recent query in our lab regarding the fixation export file, fixation.csv.

Does the frame index of the video begin with 0 or 1; in that, if the first fixation is, say, at the 4th frame, is it then the 5th or the 4th frame of the video?

We're developing a frametime critical approach in our present work, a single frame is very important in this for us as a result. Hence the basic question, but it is not written explicitly in the documentation.

Great, thank you very much for your swift support as always 🙂

The variance in frame beginning and end points in the fixations output file have somehow become inconsistent in our current data set of about 40 participants. I'll most likely follow up soon with more detailed questions which I hope will benefit everybody.

Best wishes.

In the fixations.csv file there is a column named "start_frame_index", which refers to the first frame of the video export at which the indexed fixation begins. At least so we assumed.

From what I can see the frame durations in the video are not constant, or in other words not all frames are 33.33ms periods. So calculating the frame we should target based on the duration hasn't worked for us so far. This is why we're using the frame indexes.

Unfortunately what we do right now is dependent on the world camera video, so we are constrained by the temporal resolution the video is recorded at.

Hello

Can you please provide me with a step by step instructions of how to export pupil size data from my recording in an excel format?

I used the core device to do the recording

Hello! Where can I find the audio setting in Pupil Capture? I rendered the videos, and they have no audio input or audio.timestamp files.

Thanks for your answer. Do you have any advice for further synch between audio and video files if they exported from different sources?

Our study aims to observe children’s gaze in response to partner’s verbal labels during interaction, so we need the full transcript to further identify the key timings (it’s not fixed as other experimental paradigms). Do you know any audio program can receive external triggers?

I think the easiest option (if your experimental design allows it) would be to record a clearly visible and audible "clap", e.g. with something similar like a clapperboard. Then it is just a matter of finding the clap time within the audio recording and subtracting it from the transcript timestamps. This shifts the relative time starting point from file-beginning to clap event.

Next, find the frame index during which the clap happens and extract its absolute timestamp. Add the absolute time to the relative manuscript time, shifting the manuscript time to pupil recording time.

In your pupil positions export, are your theta and phi outputs using the physics or mathematics convention? We were hoping to use your sphere radius output in combination with these two to generate a calculation for degrees of eye movements

Hi all, is there a metric to identify if the calibration went well or not? Similar to the confidence parameter metric in the csv data, after exporting with pupil player, but can be used before starting to record. Thanks in advance.

Thanks, where can I find the angular error?

Hi again, another question, can I subscribe to another topic than "gaze"? For example, if I want different information like pupil coordinates or angular error precision... If there is a list of topics to subscribe to. Thanks in advance

Yes, subscribing to other topics is possible. If you subscribe to the empty string, you will receive everything. Use it to find the data that you are interested in and then use a more specific subscription string later

Sorry I am new to this project, does this software require a infrared camera or would a regular dektop camera be sufficient to use the eye tracking technology

also hi everyone!

Ok thank you!

Hello and congratulations for the very nice work! I am interested in the pupil core system, I was wondering if it can be used with an ARM-based/embedded systems (e.g. a zedboard) instead that a normal desktop/laptop?

I see thanks for the answer - is all the source code available? Speaking about power, does the system require a GPU?

ok thanks a lot for the answers and the links I'll check them to get a better idea - it's a very interesting system!

Broadly speaking, I would like to understand how easy would it be to use it in an embedded/ARM based-system with a camera and a projecting screen

more like a wearable device - I am afraid I cannot go into more details for the moment

Hi there, From pupil service, I subscribe the frame and trying to get the image and wanna save by myself. import zmq import msgpack ctx = zmq.Context() pupil_remote = ctx.socket(zmq.REQ) ip = 'localhost' port2 = 50020 pupil_remote.connect(f'tcp://{ip}:{port2}')

pupil_remote.send_string('SUB_PORT') sub_port = pupil_remote.recv_string()

subscriber = ctx.socket(zmq.SUB) subscriber.connect(f'tcp://{ip}:{sub_port}') subscriber.subscribe('frame')

while True: array = subscriber.recv_multipart() item = msgpack.loads(array[2]) print(item) break

and I get the array which contains these 3 elements. [b'frame.eye.0', [email removed] b'\xff\xd8\xff\xc0\x00\x11\x.........................']

how can I translate this into an image or numpy array that I can save as an mp4 video? because I am receiving this error when I tried to load the 3rd element with msgpack item = msgpack.loads(array[2]) File "msgpack_unpacker.pyx", line 202, in msgpack._cmsgpack.unpackb msgpack.exceptions.ExtraData: unpack(b) received extra data.

Hi, check out this example https://github.com/pupil-labs/pupil-helpers/blob/master/python/recv_world_video_frames.py

Hi all, is there a possibility to develop the Network API with a different programming language? In C# for example.

Yes, that is possible. https://github.com/pupil-labs/hmd-eyes/ has a c# implementation of the api

Hi all, Thanks to @user-e3f20f , I can get image from Network API. I got another problem. I could not get gaze information anymore. I was receiving gaze data previously from the network by using subscriber.subscribe('gaze'). But now I am receiving the following topic: pupil.0.2d pupil.0.3d frame.world frame.eye.0 How can I get back gaze information or gaze topic? Do I need to change anything in pupil capture?

Hi, you need to calibrate to receive gaze data.

Hello, what kind of material do you use to print the camera extender?

Hi Pupil Labs community! I'm thinking to use a chin rest along with Pupil Core. Are there any commercially available chin rests that you folks recommend?

Hi, I’m getting a lot of messages like this from Pupil Capture:

2022-09-19 16:47:33,981 - eye0 - [DEBUG] video_capture.uvc_backend: Received non-monotonic timestamps from UVC! Dropping frame. Last: 19869.178274, current: 19865.838207

How can this be and is there anything that can be done about it? It seems to disrupt pupil recognition quite severely. (I’m running from source on Mac OS Monterey in native (arm64) mode.)

Thanks for your quick reply. But there is no external connection, everything is running locally on the same machine. How could a local connection become loose?

Ah, I see. Thanks, I'll check.

Hello! I am trying to refresh my questions regarding the headset extender. I found the geometry map, but not sure of the appropriate material for printing. PLA, ABS, nylon? Do you have any advice?

I recently printed it in PLA. Works nicely!

Hi! Wondering what type of machine/laptop everyone is using for the pupil core software/ what works best to avoid overloading/crashing that I’ve been noticing on my Mac laptop.

Hi guys ✌️
I have Recordings that are about 25 minutes long and try to do marker detection to extract the head-pose with the extension. The problem is, that the Pupil-Player keeps crashing. Does anyone know if there is a way to resolve it? The problem shouldnt be cpu, gpu or ram as it uses between 5-10 % on these. Have a great day!

Hey! Could you share the player.log file after having attempted the marker detection and reproducing the crash? You can find it in the pupil_player_settings folder.

Is there a windows/Linux laptop you might recommend as well ? My Mac book pro is crashing with it and I’ve checked the hardware multiple times and the logs don’t have much of any information regarding what happened. I think my Mac book pro is 2018 though with Monterey OS so could that be an issue?

the pupil player immediately gives no response

I can but it happens with every recrding after little time. The Marker detection runs for about 3-4 Minutes and then the Player gives no response

Hello, A wire to one of our pupil-cams broke, and we would like to fix it. I figured out, that I have to get it crimped. Do you maybe have the serial-numbers of the connector parts, or do you even sell replacements? Thank you.

thanks

Copying this to the correct thread: https://discord.com/channels/285728493612957698/633564003846717444/1022212302986031218

Is the stated discrepancy in accuracy/precision between VR/AR and core true in real world scenarios, if so, is there a technical reason for this?

I noticed that the pupil detection parameters in the eye windows (such as "Pupil intensity range") are not persistent across sessions, unlike the plugin parameters for the video source, for example. Is there a way to change that?

Hello everyone, when I use the pupil player, I can see a video of my gaze on the screen (in 2D). But when I export and open the gaze_position.csv file, I see a 3D gaze position (X, Y, Z). My question is: is the gaze position played on the pupil reader video simply the (Z and X) or (Z and Y) coordinates of the gaze position? Or is there a geometric transformation to translate the 3D coordinates into 2D coordinates? My goal is only to get the 2D coordinates of the gaze position. Thanks in advance 🙂

Thanks, I saw the norm_pos_* fields, but can I have for example norm_pos_0 and norm_pos_1 for the two different eyes?

Out of curiosity, may I ask about what your use case is?

Are you aware that you will need to transform the gaze position from scene camera to screen coordinates? And that you can use the built-in surface tracking feature for that?

Also, do I understand it correctly that the goal is to do that in realtime?

Ah, then I misunderstood. The visualization is for analysis only then? Not for an interaction with the subject?

Hello, -- New user here. I wanted to know if there is an existing script to stream live data into MATLAB. thanks

Hi, good catch! The issue will be fixed in the next Pupil Core release. Meanwhile, you can use this plugin to fix the issue https://gist.github.com/papr/9b3ba71b227bc6f092088077b334cf9c

I checked the headset, there are no visible defects. Also, there are never any disconnect or reconnect messages. The dropping of frames starts out of the blue for any or all of the three cameras.

It happens irregularly, but on average every 10 minutes. Sometimes it recoveres after a while, but often I have to restart.

No, many dropped frames. And when the dropping of frames stops, the cameras are (and stay) out of sync.

That I couldn’t say (hard to see what’s happening during a blink...). But I keep getting "Resetting history" messages from the blink plugin, and the red circle in the world window splits into two independent ones, presumably one for each eye.

Just realised that I missed an important detail in your initial message. The timestamp would be zero on reconnect. Yours is just inconsistent. I am fairly sure that there is something going wrong with the handling of the camera's hardware timestamps. But I can't tell what it is right now. This will require more time to investigate.

As a work around, you can switch to software timestamps. These are a bit less precise but more consistent over time.

diff --git a/pupil_src/shared_modules/video_capture/uvc_backend.py b/pupil_src/shared_modules/video_capture/uvc_backend.py
index f376eadf5..945b168cf 100644
--- a/pupil_src/shared_modules/video_capture/uvc_backend.py
+++ [email removed] -272,7 [email removed] class UVC_Source(Base_Source):
     def configure_capture(self, frame_size, frame_rate, uvc_controls):
         # Set camera defaults. Override with previous settings afterwards
         if "Pupil Cam" in self.uvc_capture.name:
-            if platform.system() == "Windows":
+            if platform.system() in ("Windows", "Darwin"):
                 # NOTE: Hardware timestamps seem to be broken on windows. Needs further
                 # investigation! Disabling for now.
                 # TODO: Find accurate offsets for different resolutions!

Thanks a lot, I’ll give it a try!

A first test indicates that the problem may be fixed completely by this workaround!

so it seems all new MacBooks Pro have M2 chip. My question is whether that may generate any troubles. M1 are still in MacBook Air, but that has 8-core CPU and is comparable with Intel i5 (maybe i7), which is listed by Pupil Labs as minimum requirements. Does anyone have any experience with M2 chips? And if not, what concrete non-Apple new laptops are people using without hiccups?

Would you say that the Mac book air with 8-core CPU is sufficient? Or should I opt for better CPU over the M1 chip?

25+ minute recordings might be likely for the experiments we plan to do. Do you have a specific recommendation of a laptop for this?

found device

Hello! I am trying to tag different windows of a software for a study. My test persons will sit between 80 and 100 cm away from the screen. Unfortunately, the tags (tag36h11) seem to have to be quite large at 260 x 260 pixels for the surface to be robustly detected. Are there ways to reduce the size of the tags and still have good tracking of the areas of interest?

Hi, when I click freeze model in Pupil Capture the red ellipse suddenly changes shape, becomes more round, and appears not to fit the pupil as well as when the model was not frozen. Is this expected behavior?

Hello everyone, is it possible to make pupil player work without drag and drop? using python parameters for example? My goal is to develop an application that uses pupil core and pupil player. Inside the application, I can: (1) launch the pupil core, (2) launch some experiments, (3) start recording (using network API), (4) stop recording (using network API),
(5) export the data. I am stuck at the step (5), because I have to quit the application and launch the pupil player to export the data.

I work on the comparison between the position of the gaze on the surface and the position of an object on the surface. I use Surface Tracker to find the coordinates on of the gaze on the surface.

Yes, I use the surface Tracking in RealTime

Hello again, is it normal to have coordinate values outside [0,0] and [1,1] for x_norm and y_norm, for the file gaze_positions_on_surface_<surface_name>.csv ?

Yes, that is when the gaze was outside of the surface. Note, it is recommended to remove low confidence gaze values as they are likely inaccurate

Hello! I'm using Pupil Core and I've been experiencing pupil detecter jumping across eye camera FOV and not staying on users' pupil correctly. I tried to change the angle/position of camera, framerate, brightness of the room, even rebooted PC and turned off one of the right/left eye camera, but they did not improve the situation. Is this a common issue with Pupil Core?

Could you create a Pupil Capture recording of you rolling the eyes (capturing different eye positions) and share it with data@pupil-labs.com for concrete feedback?

how to solve this problem？ win10

i have a usb cam

@user-e3f20f

Hello everyone, the Pupil Сapture program does not display video from glasses. Laptop model: MacBook Pro (13-inch, 2019, Four Thunderbolt 3 ports), 2.4 GHz Quad-Core Intel Core i5 processor. Glasses model: Pupil Core Please tell me what the problem might be; thank you in advance

Hi, I'm working with an older Pupil Pro Binocular rev 037 and Pupil Capture v3.5.1 but am having some issues with the setup. Cant detect both pupil cameras simultaneously, occasional program crashes when selecting camera sources, blue screen crashes. Is there some older software more suited to this hardware that I can use on a win10 machine?

You could try this process:

Windows Videos do not appear in Pupil Capture. This could mean that drivers were not automatically installed when you run Pupil Capture as administrator. You should first try to run Pupil Capture as administrator (right click pupil_capture.exe > Run as administrator). If that does not work, follow the troubleshooting steps below:

In Device Manager (System > Device Manager) View > Show Hidden Devices Expand libUSBK Usb Devices, Cameras, and Imaging Devices categories. For each Pupil Cam device (even hidden devices) click Uninstall and check the box agreeing to Delete the driver software for this device and press OK Unplug Pupil headset (if plugged in) and plug back in. Right click on pupil_capture.exe > Run as administrator. This should install drivers automatically.

found here: https://docs.pupil-labs.com/core/software/pupil-capture/

I was having issues getting my cameras to detect and this fixed it for me

Hi all! When setting up the camera an error occurs and crash program if the camera gets too close to the field of dots

Colleagues, tell me, what is the difference in calibration with the flag "show undistorted image" on and off? When calibrating, you can see that the calibration area changes.

How does raising the flag affect the calibration?

How do I start GPU acceleration for pupil capture?

i alway can't use this cam .

in win10

i can't get the image of cam

please help me @user-e3f20f

I already did what you asked, but it didn't work

If I open the app that came with Windows 10, I can get videos.

so,i think

i need your help

there are some trouble with ：Pupil Capture v3.5.1

thanks for your help. Now.i will do it as the same ways.

Could you please show the picture of the successful replacement of the driver?

@user-e3f20f

can you give me a doc or a video about this operation?

i don't know how to us this code.

i don't know how to use this code.

ok

Hi, I would like to know if any of you can help me with information to integrate PsychoPy and Pupil experiments. with LSL. UnU

not yet! Thank you ❤️

If I have any doubts, I tell you

What is the consensus on using the pupil core software with 3rd party recording devices like a web camera instead of the eye tracker?