Hello, I am a new user, and i have a quick question

I have created an acount on both the pupil labs interface and the cloud interface, how can i move the videos from the smartphone to the labtop and open the cloud from my labtop?

Hi, I have a question about the enrichment analysis. Assume I run the face mapper and the marker mapper enrichment on the same recording. Given that one person cannot physically look at both a face and an object (marker mapper) at the same time, then am I right in thinking that the enrichment data should NOT show any overlap (in timeseries) in 'true' fixations between face and marker_mapper?

Hi @user-f4b730 👋 Thanks for flagging this! Could you try opening the recording with the latest Neon Player Beta and see if the same issue occurs?

If the gaze visualisation still disappears, have a look at the Worn stream. It’s possible the worn detector was triggered and temporarily suppressed the gaze overlay.

If everything looks normal there, feel free to share the recording with us at [email removed] or open a ticket at 🛟 troubleshooting and we’ll take a closer look.

Hi @user-7413e1 👋 ! You're right that in most cases, a person can’t fixate on both a face and a marker at the same time, so you'd expect no overlap in the ⁨true⁩ fixations across enrichment sources.

That said, there are edge cases, for example, if a face appears on a screen that also has markers (like in the attached image), and both the face and marker regions are detected, you might see overlaps in the enrichment output. This doesn't mean the gaze is split, but rather that both mappers are tagging the same fixation as relevant based on their criteria.

Or, if there is some occlusion, a person stays in front of a surface and both are detected.

Let me know if you're seeing unexpected overlaps and want help digging into it!

Sure, feel free to open a ticket and if you are able to share the recording we can have a look

Would you be able to share the recording where it happens along with the timestamps, and relevant files (CSVs) at [email removed] or create a ticket 🛟 troubleshooting so we can have a look?

Hi Pupil Labs. Quick question: is there anywhere I can look for the calculation for correcting the head-movement (e.g. rotational matrices, perspective) when using marker mapper?

Hi @user-660f48 👋 ! We recently open-source the latest implementation of marker mapper, you can find it here. May I ask what is your aim, such that I can direct you to more specific?

Hi! We found this paper using relative pupil size compared to the iris size. It seems like they have only used a Pupil Labs device. However, we were not able to find the data of the iris size when we did our measurements with our Neon Module. Is this something we overlooked or does the data not provide this information (do you think this research used a second camera or something)? Thanks in advance!

Hi PL members! We have a few recordings where we forgot to turn on compute eye states, and thus they lack the blinks even on the cloud. Could you trigger blink processing on Pupil Cloud for these recordings? The recording IDs are: c17cd630-764e-4e61-87ee-95be3d6b8181 16841adb-da58-4c42-be02-f052c3c43db3 Thanks a lot in advance!

Hello PL members, I've got the following pupil Lab glasses model and I'm looking forward to get its 3D geomtry online (stl or obj file). I'd need this to help me draw and build a 3D-printed add-on but I can't find this model as of now. Thanks in advance for your help !

Let us know if you need anything else.

@user-c833f8 , one additional point:

Computing a pupil-to-iris percentage isn’t typically necessary for Pupil Core or Neon. We don’t report apparent pupil size in pixels but instead, we already provide physiological pupil size in millimetres, accounting for gaze angle and corneal angle.

I have a question, How can I export the eye tracking data to an excel sheet, including the fixation coordinates and time?

Hi @user-2501a8 , are you using Neon Player or Pupil Cloud? May I also ask if your group had its free 30-minute Onboarding call?

is this frame flexible plastic?

Hello Team, What is the storage size of the S25 included in the Neon bundle? The Neon specifications state storage is 25+ hours, but I'm wondering if I can increase recording capacity by purchasing an S25 with a different storage capacity myself.

Hello, I am using Neon with Neon PLayer 5.0.5 In one recording the player shuts down. I attach the screenshot of the cmd window. It indicates that there is a mismatch between arrays. can this be sorted out manually? or the whole recording cannot be used?

Hi @user-52e548 , the S25 we ship comes with 128GB. You could obtain a S25 that has more internal storage space if that is necessary.

Just for completeness, please note that the S25 does not have an SD card slot and SD cards would regardless not be advised. SD card storage on phones is not fast enough for Neon's data streams.

Hi @user-52e548 , just to clarify, with 128GB storage, this translates to ~16h of storage, since the Samsung OS occupies part of the 128 GB. The 25+ h refers to the standard Moto 40 Edge Pro storage. In other words, it is roughly 5GB/hr.

Hi there,

Does anyone know safe operating temperatures for the neon? I’m looking to use these during heat stress exercise. Around 33 to 38 degrees. I couldn’t find this online, sorry!

Thank you very much, will be in contact if in need! 🙂

Hi all, what is the (Working) Temperature ? Can the device work under -10 degree ?

Hi @user-af95e6 , we have not tested Pupil Core at such temperatures. It might work, but we cannot provide a definitive answer. May I ask what the experimental context is?

We want to use the eyetracking device in the frozen environment warehouse activity. The companion device could be keep under the unifrom with the heat protection case, but we are a bit worry if the camera can work actually in frozen environment or not

Since you mention Companion Device, are you using Neon, Pupil Invisible, or Pupil Core?

Its a Neon

Ok, then in the future, it would be easier to post in the 👓 neon channel, as this channel is reserved for Pupil Core, I will move the posts there and we can continue discussion in that channel.

Hi @user-af95e6 , so we have taken Neon snowboarding and skiing, where it worked great, but I think that was not below -10 C. I will check with the team to see if we have any info about how low Neon can go.

Our visualization figure is stuck in a circle and we would like to change it to crosshair. Can this be fixed

No, sorry, this bug have been going on for a couple of months. But i just tried the hard refresh and every morning im logging in and it still only shows a circle

Dear team, I am really not satisfied with the exposure of Neon. Take a look at the recording while walking today. It is unusable! The exposure is at 145 and still the scenery cannot be seen. The sun was NOT shining against the camera. Please find some solution, because Pupil Invisible are working perfect under these conditions. It is not possible that the new Neon are behaving worse. I have to say, that Balanced is even worse. There is something else: when driving on the highway and the light conditions change, it is very dangerous to change the exposure MANUALLY while driving, in order to be able to see the scenery and the QR markers. How can we work with these glasses that way? Invisible are way better, but I bought Neon in order to use them.

Hi @user-3c26e4 , could you share the full recording with data@pupil-labs.com or at least a 35-40 second screen recording around this time point, including the 20 seconds after this time point? Properly evaluating exposure for video is not really possible with a single image.

With respect to driving, have you tried our egocentric gaze mapping tool?

I sent you an email!

Hi- I recorded a video today however I do not see the recording in the list of recordings in the companion application. It says the recording saved. I tried doing some more test recordings and the same is happening, it says recording is saved but they do not populate in the app. I tried restarting the app and continued to have the same issue. Is there a way to recover these recordings?

We do not use pupil cloud because of the type of data we are working with

Yes, every purchase comes with a free 30-minute Onboarding. Simply write an email to info@pupil-labs.com to schedule it. We can cover many of these questions for you.

In the meantime, to get fixations in a CSV format that is compatible with Excel, simply right-click the recording, go to Download, and choose Timeseries CSV.

Hi can I run experiment with psychopy without wifi or internet connections?

Hello, Do I need to create the reference image and validate it before I collect data? I have static images but they will be presented in a random order. My thought is that I should create events for each participant that correspond to each image they see. But that would be after data are collected. Also, how do I link the data files with the reference image mapper to analyze data?

Thank you for your quick response! I'm not clear on how the reference image AOIs are merged with the gaze data for each participant. We have the reference images and scans. I don't see a way to apply them to data as it is being recorded. I understand that I can export the data from the Mapper afterwards.

Regarding onboarding, we are using equipment from a collaborator's lab. She has had an onboarding call but she is using the equipment for a different purpose than our work.

No, we can process it later. That's what I'm trying to figure out - how do I apply the reference image AOIs to already collected data? Or do I need to have the reference image AOIs defined ahead of time and somehow pre-loaded before starting a recording for a participant?

Ah, this is exactly what I needed to know. Thank you!

One other quick question - if static images are shown in a random order, do we create Events and label them with the image in order to analyze all images the same? For example, if my recording has images in order 1, 2, 3 and yours 2, 3, 1, is Events the way to label them for analysis later?

Yes, that makes sense. I didn't mean participant name. We are using the wearer profiles. Thank you!

Quick question. Does changing the gaze offset post-hoc (in pupil cloud) have any effect on subsequent downloads of the 3d_eye_states.csv file? I assume the eye states are "upstream" from gaze calculation but just wanted to make sure

Hi Pupil, one of our researchers encountered an issue after having recorded a session, there was no gaze data recorded: " I realized that when I processed one of the eye-tracking recordings in the Neon player, I did not get any entries in the gaze positions file. I tried exporting the recordings from the phone app and processing them in the Neon player again, but it was the same. " Is there any way to find out if the gaze data exists, and how this can be analyzed?

Hi everyone, is there a way to upload a downloaded recording (in native recording format) back to pupil cloud? or alternatively, is there a way to export pupil diameter time series of the same recording via Neon Player?

Hi @user-0f82bc , may I ask if you enabled the Eyestate timeline plugin in Neon Player before exporting the data? That is necessary in order to export pupil diameter.

Regarding re-uploading to Pupil Cloud from a PC, that is not a supported workflow. It is also not possible to re-upload recordings to Pupil Cloud, after importing them into the Neon Companion app.

Is there an advantage to the I can track clearly vs standard neon? A colleague wants to do some driving studies and I said the neons should suffice.

If you can clarify what kind of studies and what their needs and equipment are, then we can provide better feedback. For instance, will it be in a simulator or in cars outside?

I believe cars outside

Thanks for the quick response BTW

Hello @user-f43a29 , @user-d407c1 , et al. I have a question about surface tracking in Neon Player v6.0.0b3. My recordings have markers but when I add a surface and edit the surface while the video is paused, the changes do not stick and the surface reverts to the markers detected when the video is played or I uncheck the "edit" box. I would like to draw an AOI and export gaze. Thanks for your help, please let me know if I can clarify further. -Eric

Hi everyone, apologies in advance if this is not the right place. I am new to this community and have recently discovered the interesting potential of Pupil Labs products. I am an Italian PhD student working in the field of deep learning, computer vision, biometrics and human analysis, and I am looking for glasses capable of eye tracking and gaze tracking for academic, scientific and research purposes. I've explored the website but found conflicting information about who to contact (via email or video call) for further information about the products until I found the reference to this Discord server. Can you help me, please? In short, I would like to contact a specialist to ask some questions about the products to see if there is a solution suitable for my field of research. Thank you very much for your help!

Thank you very much: I found two versions of that calendar, one maybe dedicated for after-purchase support so I was a little bit confused. Thank you

I love all of the frames! I wanted to run a simulation to test the durability of the read-set-go frames, is there a CAD model available for these? I apologize if I am asking the question in the wrong place!

Hello, I recently acquired data on a subject and found considerable differences in pupil sizes between their two eyes. It is possible this is a real result, however, I want to exhaust all possibilities. Is there a way to validate the 3d eye model in Neon? I saw a thread in the "core" channel that contained an image of the pupil fit to the camera image of the eye. Does something similar exist for Neon? Thanks!

Hi @user-5cab7c , the CAD files for the Ready set go frame are not publicly available. Rather, we have open-sourced the CAD files for the Just act natural frame.

What kind of simulation would you be running? In what situations do you plan to use Neon?

Hi @user-429d6a , would you be able to share the recording with [email removed] Then, we can take a closer look.

Ah, do you mean Airsoft? In that situation, you could also consider the following frames:

• Safety on
• All clear

If you prefer Ready set go because of the headstrap, we also offer a headstrap for the Safety on.

I can ask the team about impact ratings for the frames specifically.

Ok, I will update you when I have a response from the team.

Thank you!

Hello. I am interested in purchasing an academic license for pupil labs cloud storage. What is the best way to do that?

Hello @user-f43a29 , @user-d407c1 , I am using the Neon API to save session data to my local PC. There is consistently a ~4–5 s (not ms!) latency (the local data files are a couple of seconds shorter than the data saved to the S9 tablet). Is there a way to address this? Also, the data saved to the tablet are UUID-named, and when translated to local time, they also have timestamp offsets relative to the folders with script-generated timestamps when the API controls the Neon App to start recording. The timestamp offsets range from a couple of seconds to +60 s (I have checked the scene camera and confirmed that they refer to the same session). Is there a way to systematically align the data saved to the tablet with the ones on the local PC? Thank you!

Hi @user-d4c059 , no need to worry about tagging us, as we will see your message.

First, may I ask your use case? To clarify, the S9 tablet is still an experimental device.

Next, may I ask why you are saving the streamed data rather than running a standard recoding in the app? The recorded data will be higher fidelity and will be in a format that is compatible with the tools in our ecosystem.

Once that is clarified, we can proceed to the timestamp offset question.

@user-d4c059 , you are welcome and nothing to apologize about.

How long are your sessions? You can record for 4 hours on a single charge of the phone. If that is not enough, you can use a USB hub to simultaneously charge the phone.

Note that it is not possible to start/stop Neon's sensors at the same time, nor in synchrony with another device, even when scripting. Neon's sensors all have non-deterministic start times and run in parallel. It is instead recommended to use Events and follow our Time Sync guide or use the Lab Streaming Layer integration.

If you are only working with data from the phone, then make sure to use the Export button in the three button menu next to each recording in the Recording List of the app. Then, it will not have a UUID, but rather the date and time at which the recording was made. You can then rename the folder after copying it to your computer's harddrive. You could even automate this with the adb tool.

Hello all, I'm working with a separate developer to use data recorded in the Neon Companion App from the Neon Module for features in a Unity project. He did not use the Unity XR Package and instead wrote his own system. Data Transfer was working just fine for weeks now but out of nowhere, specifically tracking data was no longer being transferred. We can still access the camera data but nothing is being received in terms of pupil tracking. I've already restarted both the computer and phone as well as ensuring they are on the same wifi network. The Neon App also displays that there is one device connected when I have Unity running so I know that it is able to transfer some data to Unity. There should also be nothing blocking this data transfer in the computer's firewall settings. Is there a solution to this? I know the tracking is working in-app so I think the issue is in the app not the hardware.

Dear Pupil, in the .csv tables there is only Recording ID, but the names of the subjects are in Sections.csv. Is it possible to download for axemple aoi_metrics with both Recording ID and Wearer Names in colomns?

Hi @user-3c26e4 ! Feel free to suggest it on the 💡 features-requests channel. In the meantime, you can add You can append the wearer's name to any CSV file with recording id using this snippet. https://gist.github.com/mikelgg93/487abeb80fd26ff90a101189250b1357 Lemme know if you have any trouble using it.

Hi @user-f43a29 We had in our last project some issues with detection of blinks. Our Companion Device in combination with the actual neon monitor sometimes does not detect blinks correctly (we are obliged to use the system locally and we are not allowed to use cloud-storage). If I would use the asyncAPI (python) receive_eye_events_data method will it be computed seperately out of raw data from the NEON device or does the method rely on the data precomputed by the companion device?

Hi, could you please confirm whether this method for calculating the horizontal eye position (in degrees) using the optical axis is correct?

df["azimuth_left_deg"] = np.rad2deg( np.arctan2(df["optical axis left x"], df["optical axis left z"]) )

df["azimuth_right_deg"] = np.rad2deg( np.arctan2(df["optical axis right x"], df["optical axis right z"]) )

Hi @user-f43a29 and @user-d407c1 , We've been using neon successfully with python and are trying to deploy it with some code written in MATLAB (R2023b running on Ubuntu). We've followed the directions vis a vis the support matrix (Python 3.10, appropriate library versions etc.) and configured the MATLAB variables (pyenv etc.) carefully and correctly (as best as we can tell) to enable matlab to talk to the glasses via python. However, we're unable to discover the device via MATLAB, i.e., run device=Device(); successfully, even as we continue to be able to talk to it via python) After some debugging, we've narrowed down toDevice.m's obj.py_device.receive_matched_scene_video_frame_and_gaze(); on line 60 upon execution of which MATLAB hangs (stuck on busy state) and remains unresponsive. Although this seems to suggest an inability to communicate with the device (network issues), we are able to exercise it fully via existing python code for a different experiment. We also were able to retrieve device IP etc. Do you have any advice or suggestions you could offer us on where we could start looking for root causes for this issue?

Note: I'd initially written with a different set of detailed symptoms which I was able to resolve and hit the specific problem mentioned above. If you read that earlier, please excuse the edit. Thank you for your help with this.

Hi everyone, I am interested in using the Every Move You Make version of the Neon Glasses to get the absolute position of the pupils and gaze. Is the tracking of the marker already integrated into the software, or do I need something else for that? Are there cameras recommended for tracking the markers? Thank you 🙂

Hi @user-27f0ef @user-97c46e 👋 ! You may want to check the latest update in 📯 announcements , including the compatibility notes.

@user-27f0ef since you're using a custom solution rather than NeonXR, you can remain on the previous app version until your software is updated to support the changes.

@user-97c46e you may simply need to update the underlying Python realtime library to restore compatibility.

Let us know if anything remains unclear after checking.

Thanks you for your response. Could you say a bit more about what you mean by restore compatibility? I am able to use the device via Python already, without any issues. The compatibility requirements for MATLAB interop (https://github.com/pupil-labs/pl-neon-matlab) are also met. Is there a interop matrix between MATLAB/OS and the python library version that I'm missing based on which you believe updating the python real time library is the missing piece?

Thank you for the quick reply. By "absolute position," I meant: For a project, I need to know the exact location of the driver's eyes and pupils in 3D space. I just started working on this and am not sure if there is a simpler solution or better wording 🙂 .

Do you have a recommendation for a motion capture system that works well with python or would it be possible to have a scene camera that is fixed to the car?

We own the one+ 10 Pro, (NE2213) with oxigenOS 13.1, neon companion 2.9.8-prod. The problem was: some of our recordings did not detect blinks correctly. Our workflow was: after a day or so of recordings (in traffic simlator, ~3h of driving) we exported the raw files to a disk and deleted the raw files to avoid storage issues on the device (70 persons, ~1h recordings per person). But we saw afterwards, that some of the recordings seems to lack the blinks. the final solution was: pupillabs was so kind to recalculate the blinks after uploading the files to the cloud. This was OK for the moment as the iris and other data out of the video were sufficient pseudonymized. The scene camera does not catch persons, nor real scenes. If the pre-computing is on the device, then programming will not change the game. to avoid this: Can we connect i.e. a windows / unix machine instead of the companion device? According to my idea, more calculating power = more accurate detection blinks offline, am I right?

@user-d407c1 Hi Miguel, Do you have any information on whether it is possible to identify microsaccades with pupil neon glasses?

Hi @user-d407c1 , We are planning to collect new data. but they should be comparable to the 'old' ones regarding fixations / saccades and - of course blinks. I will update the app.

one question still not answered. is it possible to use a unix/win-machine as companion device? As we are most of the time in door at a driving simulator it could be more comfortable?

Hi, I have a question regarding recordings. I wanna do recordings locally and display pupil data on the experiment screen in real time. Is it possible?

Thank you! Could you point me to resources on how to integrate a fixed scene camera, or suggest someone I could contact? I really appreciate the good support, and I'm looking forward to using the glasses!

Thank you for your answer. We own as a companion device the (maybe early) one+ 10 Pro, (NE2213). Is it known to the one+ to struggle in 200Hz-detection? In short: If we would like to integrate (=get the data of NEON and the simulator synchronous in one file) data , we shold use the WLAN-extension of the sim-network and use the async API, knowing that 200Hz is sometimes 'too much' for the One+. We usually conduct/sync the simulation by 120 Hz.

Hi! We are moving our analysis away from the cloud and using Neon player, but I'm now experiencing problems with loading in the data. I have exported from the phone and have the folder, and I can see video/audio/gaze in Neon Player fine. But when I try to detect fixations using the plugin it takes a while, percentage ticks up and then it detects zero fixations

Good morning/afternoon, I am trying to download the "native recording data" from pupil cloud Neon recordings so I can process them in Neon Player. I have noticed that some sessions, despite working fine in the cloud, whenever I download them, my macbook throws an error 79 "inappropriate file type of format".

I'm wondering if this is just a mac problem? I noticed that I also couldn't install Neon player on my macbook but I was able to run it on my windows laptop. I was only trying to fetch the recordings and copy them over to an external hard drive but the error had me worried there's something wrong with the recording data.

[email removed] Byron , could you try opening

Yep I am using psychopy and I wanna do things as you mentioned