Hi, how do I get the latest and greatest version of pupil capture WITHOUT having to run from source? You last release was in 2021, but there are way newer commits, even a few months ago, e.g. on the develop branch.

EDIT: keep up the good work!

Hi @user-84387e , it is possible to bundle it locally. Do I understand correctly that this was already tried? And thanks!

No intentional trickery. We use pyinstaller, like you said, but then we use WiX to wrap that up in a Windows installer (despite the name, pyinstaller does not do that itself).

You can install WiX yourself and give it a shot, although you'll want to run the reproducible_build.ps1 script in PowerShell (rather than the .sh).

However, you may find it simpler/easier to whitelist the folder so Windows doesn't scan it for threats. Search of "exclusion" on this page: https://support.microsoft.com/en-us/windows/virus-and-threat-protection-in-the-windows-security-app-1362f4cd-d71a-b52a-0b66-c2820032b65e

Hello!

I'm experiencing an issue with the Pupil Labs app while annotating data. Previously, I could add multiple events without any problems, but now the app stops responding when I try to add more than seven. Once I reach seven events, I can no longer click on the sidebar where the events are listed. Is this a known issue, or is there a way to resolve it? I already tried restarting, and checked the updates, there's nothing wrong with that.

Hi ! Could I use https://github.com/pupil-labs/gaze-controlled-cursor-demo?tab=readme-ov-file with my Pupil Core ? My device is not detected

Thanks

Good evening folks! I was looking through some of the archived threads about Neon and the Core Pupil Capture platform but didn't see this specific issue. I did a big data collection for my research using pupil core but we had some issues with participants not being able to see without glasses. I ideally need to have these participants perform my experiments and we have Neon with the corrective lenses but I figured it would open my research up to scrutiny if some of the participants were recorded with a completely different eye tracking platform. However I thought a potential work around is using the Neon headset with Pupil Capture.

The problem I'm running into though is so far I cannot get it to work at all. I am able to get Pupil Capture to recognize the world camera of Neon, but the eye cameras are not recognized and this is confirmed when I attempt to calibrate and it says "Not sufficient pupil data available". When I was working with adding Neon as the local usb camera, I see a second option for Neon Module, however if I select this, it immediately crashes Pupil Capture and my laptop to the point I have to restart it. I understand it is experimental but I was hoping to at least get a recording going. I'm running a Macbook Air, I can potentially arrange to test it on a windows machine as well but was wondering if there were any suggestions or if anyone has run into this before? Thanks!

Hello! I am working with the Pupil Labs fork of pyuvc to control the Pupil Core arm camera, and I have observed the following behavior when exposing the camera to 0%, 10%, and 100% light intensity using an industry light generation device. Due to this behavior, the camera maxes out when in sunny outdoor settings, as this is with the IR emitters on the camera itself blocked. I have also set the camera to the following Auto Exposure Mode. I would expect the AGC in the following plot, if it were working properly, to attempt to push the signal to the midpoint of the value range? Any thoughts on how I could get it to do that in this example as well as outdoors?

I don't have a graph exactly, and I can get the luminance measure if needed, but off the top of my head I know the illuminance is roughy 0/300/3000 lux at the 0%, 10%, 100% levels. Essentially, I set a level, kept it at that period for roughly 30 seconds, then went to the next level

Oh! Sorry, I misunderstood your question too. The graph above (Pupil Mean Pixel Intensity over Time) is what the pupil camera measured. It's the average pixel intensity per frame over the recording period. I would expect, if the AGC were working properly, that during the periods where the pixel average is extremely low that it would attempt to push those higher, and vice versa when too high

Yes, I wrote pupil_utils. Basically, that variable is just converting a boolean TRUE/FALSE to the integer representation. I saw an integer used to select mode from the following file on the Pupil Labs Github.: https://github.com/pupil-labs/pyuvc/blob/master/examples/access_100_frames.py. I'll take a look at the commit you sent, thanks! And yes! Sorry, I should have clarified. I am using AGC to stand for automatic gain control

Thanks for sending the link! With a cursory glance, it seems that AEC (auto exposure control, also valid as you say for the behavior I want) is enabled by default and with similar configurations to the control dict (though, I am missing a few, like the priority for instance). Any thoughts on that being the case and seeing no real adjustment happening in regards to the plot? I will give it a more in depth look and try some things out as well

Sweet thanks!

Hi, I am trying to write an C# Code to connect to my Glasses. The HTTP REST Calls to get the status and start recording are working fine. But when I try to connect to the WebSocket I always get the Error 'Connection'-Headerwert 'Upgrade,Keep-Alive' is not correct. Has anyone suggestions for me? Thanks

Hey pupil labs, I have an issue with running the Google collab code for "Map gaze into an alternative egocentric video"

Iam getting an error as this StopIteration Traceback (most recent call last) <ipython-input-11-2cb29c821ee6> in <cell line: 0>() 26 args = Args() 27 ---> 28 main.main(args)

1 frames /usr/local/lib/python3.11/dist-packages/pupil_labs/egocentric_video_mapper/optic_flow.py in calculate_optic_flow(neon_timeseries_dir, alternative_video_path, output_dir, optic_flow_method) 266 optic_flow_method="farneback", 267 ): --> 268 neon_video_path = next(Path(neon_timeseries_dir).rglob("*.mp4")) 269 optic_flow_method = optic_flow_method.lower() 270 Path(output_dir).mkdir(parents=True, exist_ok=True)

StopIteration:

It says it's not able to find the mp4 file inside the folder.

How do I resolve this issue ?

Hi @user-b6f43d! Obvious question, but is the .mp4 inside the folder?

Yeahh, i just uploaded the whole time series + scene video folder. Didn't make any changes.

https://drive.google.com/drive/folders/1hCHzXYmMA04mxNZZX8AotM4G9AYkSw0C?usp=sharing

Thanks Rob. I uploaded the files directly in Google collab and it's works.

But again Iam getting a error saying there is some error is the video file it is trying to decide.

It says, error splitting the input into NAL units.

Hi @user-b6f43d , would you be able to put your Neon and Insta360 recording on Google Drive and share it with [email removed]

I just sent an email, please check. Thank you.

Hi, I'm currently working with the Pupil Labs Neon device to record video data, and I understand that it includes an integrated IMU sensor. I'm exploring the possibility of using this device for running SLAM algorithms. I wanted to check if your team has conducted any trials or evaluations using SLAM (e.g., ORB-SLAM3 or VINS-Mono) with this device, as I’ve tested both but the performance hasn’t quite met expectations. Additionally, I’d really appreciate it if you could provide the IMU noise parameters for the device, such as: Accelerometer measurement noise standard deviation, Gyroscope measurement noise standard deviation, Accelerometer bias random walk noise, Gyroscope bias random walk noise, I guess these would help a lot in tuning and optimizing the SLAM pipeline. Thanks in advance for your help!

Hi, when I use Neon, I notice some errors on the gaze position which depends on who uses it. Some people’s gaze positions tend to be on the left of the actual positions while others may be on the right or above. So I’m wondering is there a calibration that can be done to mitigate this issue when Neon is used across different users? Thank you!

I mainly looked into the screen-mapped gaze from the real-time screen-gaze package, but I think it's both in terms of screen-mapped gaze and gaze estimates as they appear in the Companion app. I've only checked with the center of the circle in the Companion app's video stream.

Handling pupil core data recorded from different applications.