where is it
Hi @user-1811d7 , I am not sure I understand. What are you looking for?
Hey yall, I'm trying to run pupil capture with a diy headset, but I keep running into this error. Any Advice?
Hi @user-8f6642 - thanks for reaching out! As a first step, could you try restarting Pupil Capture with default settings? This button is found in the General Settings tab.
Please note that while we are unable to provide dedicated support for Pupil DIY projects, other members of the community who have built a DIY headset might be able to help.
εθη₯θ― 40/5000 Reference Knowledge: 40/5000 AI Translation and Mapping AI Large Model Translation Please provide me with the Pupil_LSL_Relay plugin for core, as the download link on the official website is no longer valid
εθ―
Please provide me with the Pupil_LSL_Relay plugin for core, as the download link on the official website is no longer valid
Hi @user-906e58 , are you referring to this link? It is still up and valid.
well um i seen a product by pupil that was eye tracking for the vive pro 2 and i dont know were to buy it
Our original eyetracker, Pupil Core, has an adaptation for the Vive Pro, but we do not have any products for the Vive Pro 2. May I ask where you found that information?
um i seen this video https://youtu.be/eVsNdQL7VEI?si=eTliQQAYP5OY7JgL
tho it does say vive pro could that not fit the pro 2?
Ah, no, it does not fit the Vive Pro 2. If you would like to inquire about the Vive Pro add-on regardless, then please send an email to [email removed]
and were could i even get this
oh well i apologize ik i sound a bit slow, you think there will ever be any for the pro 2?
Nothing to apologize about and you do not sound slow!
We currently have no plans to make a mount for the Vive Pro 2, but you could 3D print your own Vive Pro 2 mount for our latest eyetracker, Neon, which is modular.
good to know, thank you for the help I'm gonna probably look into that
Feel free to ask questions if anything is unclear.
Hi @user-7c9706 , there is no need to re-install. You can change the port in the Network Plugin settings.
That won't work, as I can't even open Pupil Capture. It gets stuck on the interface shown in the screenshot below until the software crashes.
Can you try the following:
pupil_capture_settings folder in C:\Users\<username>\thank you for your help. It worked.
Thank you very much for all your help! I have successfully implemented wireless transmission for Pupil Core using a Raspberry Pi as a relay.
This is all based on the GitHub project: https://github.com/Lifestohack/pupil-video-backend . However, some parts of the code need modifications, and I will provide detailed instructions for those changes in the issues section of the project (my GitHub ID: away0730).
That said, I still have a few issues that I need help with:
1.I need to use the calibration feature in Pupil Capture for gaze position mapping, but due to the performance limitations of the Raspberry Pi and the wireless network transmission rate, I can only transmit at a resolution of 320x320 for the world camera. At this resolution, the camera is unable to recognize the calibration points on the screen. Do you have any viable calibration methods for low resolutions? If not, what resolution would be required to make calibration possible? 2.Due to the transmission speed, there is a significant delay when the video reaches Pupil Capture. Does the software automatically perform time synchronization to align the frames?
I would greatly appreciate your help!
Hi @user-7c9706 , great to hear about your progress!
Yes, 320x320 is quite a low resolution and it is unlikely you will be able to calibrate that way. You could try the Natural Features Calibration Choreography and see if that improves it, but it does require good communication between the wearer and the operator.
With respect to time synchronization in this case, Pupil Capture itself will not account for the network transmission delay of the videos, but perhaps the pupil-video-backend has code in it to handle that. I recommend reaching out to the creator of that plugin.
In any event, provided the network connection is relatively stable, then all frames should in principle roughly experience the same delay, meaning that your world video and gaze data could potentially be post-hoc aligned by a constant offset. You could measure this delay/offset by shining an IR light into the eye cameras and measuring how long it takes for that to show up on the recording computer.
Good afternoon everyone, how are you all?
I am a complete novice when it comes to programming languages; I am learning now because of a scientific initiation project at my university, where I am using glasses for a study with healthcare professionals.
I would like to know what the first step is to obtain information about the results of the video data collection. Is it possible to obtain it using only CSV files, or is Python also an option?
Finally, is there any article or other supporting data that provides an average for each result? This would allow us to observe changes during the research period.
Thank you in advance for your attention.
Hi @user-b11a92 , welcome!
May I ask which eyetracker you have?
Hi! I have Pupil Core.
Ok, thanks. It is currently after-hours for the team, so we can provide more support tomorrow, but briefly:
Feel free to reach out if you have other questions.
Thanks for your helpοΌ
Based on your official website, there are various algorithms available for gaze tracking, and your latest generation of eye tracker, the Neon, uses the Neon Net algorithm. Could you please confirm which algorithm is used on the Pupil Core?
Also, I noticed that LattePanda seems to natively support your Pupil Capture software. Does this microcomputer have enough processing power for gaze analysis? Are there any successful cases using LattePanda for this purpose?
You are welcome!
So, Pupil Core uses more traditional dark pupil detection with calibration, bolstered by the pye3d model.
Although we know that the LattePanda can natively run Pupil Capture, we do not have experience with this platform ourselves, so I do not know of its performance or of published cases. Perhaps someone in the community here can chime in π
Hi everyone, Iβm designing a usability study using Pupil Core, and Iβd like to combine eye-tracking data with behavioral performance metrics in a task-based setting. Specifically, participants will interact with a clinical data entry system, and I want to capture gaze data (e.g., fixations, scanpaths) alongside metrics such as task completion time, task success, and possibly interaction events (clicks, inputs). Are there recommended tools, frameworks, or software for recording and exporting these performance metrics in parallel with Pupil Capture?
Hi @user-74b1c6 , while we don't produce any software for recording such metrics, there is in principle nothing stopping you from using Pupil Core in parallel with a performance collection software of your choice.
With respect to our tools, you can at least use the Surface Tracking functionality of Pupil Core (either in real-time or post-hoc) to map fixations to your screen and thereafter, produce a scanpath. I recommend checking the other tutorials in that repository, if you are new to analyzing Pupil Core data.
Perhaps a member of the community can provide advice for performance tracking software based on their experience. Otherwise, feel free to describe your setup in more detail and we could provide some tips. For example, are you comfortable working with Python and does the data entry system have an API?
Hello, hope you're doing well. I'm reaching to you regarding Pupil Core for children. Is there a child version for the age range of 8-12 years?
Hi @user-4d0d35 , the standard Pupil Core will probably fit that age range, but if you would prefer the child sized frame, then please reach out to info@pupil-labs.com for more details. I also recommend looking into our latest eyetracker, Neon, which is calibration-free, fully mobile, and features frames for children and adults/teenagers.
Hello all, I have a few recordings, which have extremely erratic eye tracking, the primary reason being that the 3d model tracked the 2d eye positions sort of okayish, but the 3d model detection did not do a good job. Is there a way to recompute the eye gazes?
Hi @user-3d30b4 , yes, there is. Check the Documentation here and here. You can wait until a better moment in the recording, pause pupil detection, reset the 3D eye model, and then resume detection. Let us know if you have any questions.
Hi @user-f43a29 . Thanks for the quick reply! I tried doing this by myself, where I can only see the post-hoc-detection. My videos do not have a calibration part. I was given these videos, where they said that the calibration was performed for this video before it was recorded. The post-hoc detection still messes up the detection of the pupil centers. Is there something I am missing here?
You can share the recording with us [email removed] and we can take a closer look.
I have emailed it!
Thank you so much for your enlightenment; so there is only one child version of Pupil Core!
Correct!
Hi, Iβm using the Pupil Core on a 2023 MacBook Pro. Iβm a complete beginner, so I apologise if this is a simple question.
I set up in Pupil Capture and have the surface tracker plugin installed. In the recording, I set up show heat map. When Iβm done recording, I open things up in Pupil player and select show heat map under the surface tracker section, and one of two things happens.
It shows the heat map, and then as the video proceeds, it stops showing it, usually after a few seconds.
It doesnβt show the heat map
Any suggestions as to what might be happening?
Hi @user-bed841 , it potentially sounds like the AprilTags are not being robustly detected, but without seeing the recording, it is hard to be sure. Would you be open to sharing it with data@pupil-labs.com and we can take a closer look?
Of Course, thank you.
Hi Rob, I just sent them with a few other Q's if that's okay.
It is currently after-hours for us, but we will take a look and get back to you tomorrow.
Of course, no rush.
Hi @user-bed841 , thanks for sending. I took a look. Here is my feedback:
Thanks Rob! I'll reset things and give it another go, and I'll go through the best practices and initial fitting. I had done that before, but it's been a few weeks.
Hello!
I am analysing data extracted from Pupil Core, and I was wondering whether the system always samples and records data such as pupil diameter, pupil position, timestamp, etc., or if, in cases where the pupil is not detected, no data is written or registered.
I have noticed that in my data, when looking at the difference between consecutive timestamps, there are gaps of up to 4 seconds between one datapoint and the next. I am wondering whether this could be due to limited laptop capacity (we were using a laptop with 8 GB of RAM), which may have caused issues during data writing, or if it is because during those 4 seconds the pupil was not detected and therefore no pupil data was recorded for that period.
Thank you very much in advance for the clarification π
Hi @user-d03324 , what you described does not sound like a problem with your laptop. Rather, it might be due to positioning of the cameras.
If you could share a screenshot or send the recording to [email removed] then we can provide more direct feedback.
Hi @user-f43a29 - Your suggestions work. One odd thing, the heat map records and shows up in the player, but it freezes. So if I jump to frame fifty, it'll display where the heatmap was, and when I hit play, it just stays there. Is this a processing issue on my comp or have you seen this before?
Hi @user-bed841 , you are welcome. May I ask for clarification:
Hey @user-f43a29 I might have figured it out, in the player I lowered the april tag parameters and turned off Sharpen Image and Use High Resolution and then it started to adjust more.
I meant that the heatmap doesn't move over time dynamically. At the moment, after the changes I made, it holds and then snaps to a new position every so often.
Thanks for the clarification. So, the heatmap visualization is actually not supposed to dynamically update over time. It is a aggregate of the data across the whole recording and therefore static. Pupil Player does not produce dynamic heatmaps.
The changes that you made will actually reduce AprilTag detection accuracy in most cases and thereby the accuracy of the heatmap, hence the snapping jumps. It would be advised to return to the default settings.
Do you specifically need a dynamic heatmap for your research? Which eye tracking metrics in particular do you want?