Is there a way to bypass the connection requirement or trick the companion device to have an IP? My goal for my app is to be self contained and used on the go without a dedicated connection.

Ideally not. Are you suggesting using a hotspot?

The context is that we would want to run this data viz program out in the field, and we want to minimize the number of components.

Hi @user-d801e5 , the team is now aware of your request. We could have a response to you by next week about this.

Hi, I had a question. I am using the pupil labs spftware. more specifically pupill capture. I am using a raspberry pi and I am trying to stream the feeds to pupil capture, but I am running into an issue where pupil capture is unable to detect my cameras. I was wondering if you had any suggestions on how I can do this. Thank you!

Short question, how is the boundary box for the face calculated for the face enrichment? There is a p1 x and p1 y for the starting point (which equals which one? right top?) and the ending point p2 x and p2 y [px]. What point sdo they equal? Does the enrichment just assumes symmetry? So no facial marks for the boundary box and only for the facial landmarks such as eyes/mouth/nose?

@user-480f4c I simply wanted to plot the face coordinates myself using a python code: plotting the face boundary. and then the eye, mouth and nose coordinates within

thanks for clarifying @user-5ab4f5! This should be straightforward. In fact, we have a tutorial on our Alpha Lab page that maps gaze data on individual facial landmarks and draws the bounding boxes around the detected faces, along with areas of interest around each facial landmark: https://docs.pupil-labs.com/alpha-lab/gaze-on-face/

Specifically, for rendering the box around the face, this function might be helpful: https://github.com/pupil-labs/gaze-on-facial-landmarks/blob/main/src/pupil_labs/gaze_on_facial_landmarks/map_on_landmarks.py#L15

Thank you so much @user-480f4c !

Hey as an update for this, I was reading the under the hood doc, which says "The necessary connection information is made available via the Sensor model as part of the Get Current Status and Websocket API," regarding the streaming API. Is there a code example of how to do this (i.e. without the python package)?

I'm happy to stream over any protocol, I just want to understand how the API is operating so I can use it in Kotlin.

I have been trying to use the okhttp package to connect to the API through websocket, but it seems I need more than just that to stream the data according to the documentation

Happy to provide more details

Okay thank you I will take a look

This might be a silly question but I am having trouble with path(s) in /api I am interested in. I think /api/status is relevant, but not sure how to use what it's returning (https://pupil-labs.github.io/realtime-network-api/#/status/get_status)

[email removed] , could you share the code

Hello, I'm making good use of your core product—thank you for providing it.

I’m writing to report an issue related to using your open-source Pupil code with pyinstaller to generate a .exe file.

Our goal is to send the gaze point (x, y, z) vector values via Windows Messages and receive them back to be used in a display module. The main Python script runs perfectly fine in our virtual environment, where the glfw module is properly installed using pip.

However, once we build the .exe file with pyinstaller, running it results in an error stating that the glfw module cannot be imported. It seems the .exe file cannot locate the glfw module, even though it works without issues in the Python environment.

Since the team working on the display module does not have experience with software development, we need to deliver the functionality as a .exe file. This is for academic research purposes in a graduate school setting.

We would really appreciate your kind help and advice on resolving this issue.

Best regards,

@user-fce73e , my colleague, @user-cdcab0 , has reminded me that you can also simply make a Pupil Capture plugin to send the Windows Messages. Then, development should be easier and you can use the default, bundled EXEs on our webpage, without needing to deal with WiX or pyinstaller.

Hello! I am running into this error: ImportError: cannot import name 'builder' from 'google.protobuf.internal' (/usr/lib/python3/dist-packages/google/protobuf/internal/init.py) [ros2run]: Process exited with failure 1 In the past, upgrading protobuf worked: ' pip install --upgrade "protobuf>=3.20,<5"'. However, this time, when I followed the same step, the error persisted. Any suggestions?

Yup, this is to use the Real-time API. I uninstalled and reinstalled, but I'm still experiencing the same issue. My Python version is 3.12.9

I was using conda with ROS2, and apparently, that caused an issue, although I was able to locate protobuf while being in the conda environment. Deactivating the conda env seemed to make it work.

Yes I have resolved the issue by deactivating the Conda environment!

Hi there, I'm using Neon Player v5.0.4 to process data collected with Pupil Labs Neon child glasses, and we’ve encountered several issues that prevent us from opening certain sessions. Here’s a breakdown: 1. Corrupted Scene Camera Video File Causing Crash Some folders contain a corrupted Neon Scene Camera v1 ps1.mp4 file, which crashes Neon Player upon opening. Error message: Invalid data found when processing input: 'F:\03140324\2025-03-22\T16\neon\2025-03-22-18-04-27\Neon Scene Camera v1 ps1.mp4'; last error log: [mov,mp4,m4a,3gp,3g2,mj2] moov atom not found This file cannot be opened in any video player either. Is there any recovery method for this file, or a workaround to process the session without it?

2.eye0.mp4 File Too Large — Neon Player Crashes In the folder: 2025-03-15-19-01-20\neonplayer\eye0.mp4, the file size seems to be extremely large, and Neon Player crashes instantly when attempting to open the folder. Is there a known file size limit? How should we handle this?

1. AttributeError on Scene Camera Matrix — File Won’t Open One folder fails to open with this error: AttributeError: 'NoneType' object has no attribute 'scene camera matrix' What does this error indicate? Is it related to missing calibration or corrupt metadata?

2. ValueError: cannot convert float NaN to integer — Crash at the End of Download During the download of one session, Neon Player crashes near the end with: ValueError: cannot convert float NaN to integer This seems to happen after all video and data files are nearly done downloading. Is this a known bug?

Let me know if you need access to the specific folders or logs — I’d be happy to share more details. Thank you in advance for your support!

Hi, @user-03bf4a - sorry to here you're having some troubles with your recordings. My responses are below

1. (Corrupt scene videos) Sometimes you can recover videos with this type of error, but it's a real dice-roll. There is a tool called "untrunc" that we have built upon for this process. Good luck!
2. (Extremely large eye video) How big are we talking? You might be able to use ffmpeg to re-encode the video (and thus, hopefully reducing its size), but if you're seeing large eye videos on many recordings we may want to troubleshoot the cause
3. (no attribute 'scene camera matrix error) It could be missing or corrupt data, but it's hard to say. We'd probably need the full log and/or a copy of your recording to troubleshoot. I have never encountered this error myself.
4. (Crash at end of download) When you write "downloading", are you referring to the export process? This would be the first report of something like that, and we'd need a copy of the recording to troubleshoot further.

If your recordings are on Pupil Cloud, you can invite [email removed] to your workspace and let us know the recording IDs we should be looking at. Otherwise, you can upload any recordings you'd like to share to a file sharing service of your choice and share them with [email removed]

I asked this in the thread I opened, but maybe someone else can help too:

Hi, so I've gotten the RTSP streaming up, but only when connected to a network. I believe the network requirement can be bypassed by turning on the device's hotspot, but when I do that, the companion app just says waiting for dns service. Am I doing something wrong? Do I need to enable anything?

Pupil Capture in Ubuntu with multiple screen display

Saving the default settings on pupil capture for the experiment