Hi, I wonder is there a way to drag the recording video in pupil cloud in millisecond precision to achieve sync with some stimuli appearing in the video, thanks! Also is there a way to directly show UTC timestamp on the pupil cloud?

Hi @marc , thank you for your reply! So may I assume the systemic bias of this kind of sync approach is 30 ms?

Yeah you're right, thank you! About the UTC timestamp, so the first timestamp in the world_timestamps.csv corresponds to the recording.begin in the pupil cloud, right?

I see, thank you!

Hi! In gaze.csv, what is azimuth and elevation?

Hello, I would ask a question about post-hoc gaze calibration. These are my steps: I record my data from a Raspberry pi, then post-hoc the data via a desktop PC. I calibrate the eye tracker on the Raspberry pi and unchecked the pupil detection during the recording. Then, on the desktop side, I select post-hoc pupil detection. My situation is: when I only select Post-Hoc Gaze Calibration, I do not have data on gaze direction. I only get data on gaze direction when I press the button Detect References (see figure bellow). And my question is: is it the right way for having the gaze direction data? Or do I need to do other steps? Thank you in advance in your instruction, have a nice day.

Good morning pupil Labs! I have a short question, how can I download the video with the red Circle from Pupil Cloud to my pc?

Ahh wonderful! Can I set the starting time of the enrichment (the red circle) be at the start of the lsl labrecorder ?

By default on the Events it has a lsl.time_synch.etc, is this the right event, or i should follow the lsl time alignment procedure to get the json file and then create the corresponding event ?

This one : https://pupil-invisible-lsl-relay.readthedocs.io/en/stable/guides/time_alignment.html#time-alignment-under-the-hood

Ahh, so this event with red arrow, is the beginning of the LSL! That means that is the beginning of the LabRecorder? Because the LabRecorder works with LSL!

Hey, I know this gets asked many times, but we are having issues reliably connecting to the pi.local:8080 and pi-1.local:8080. We can usually connect to either the pi.local or the ip address without issues the first time, but we if close the web browser and try to open the link back up, we just get a white screen.

We are not using an institutional router and we've reserved the IP address so we could create a url shortcut on our desktop to quickly open it during our study

@nmt I've tried refreshing the browser and it just stays white. It's not even just from closing the browser, it's refreshing the browser, closing it, etc. We work primarily with kids, so sometimes things come unplugged and we are just trying to understand how we can reliably get the monitoring to work (should we need to troubleshoot/close something).

For example, yesterday in testing I got both monitoring apps to work but after closing out the browsers, to test it again, I couldn't pull up the browser interface. The webpage just stayed white. It only worked if I opened up chrome (for the first time) and tried, but then even that stopped working after closing that browser.

Monitor app streaming

Hi Pupil Labs Team! I saw that its possible to buy an "Pupil Invisible Companion" package from your website. I was just wondering: What does it include in comparison to when I simply buy an OnePlus 8 Smartphone myself?

Hello. We share an Invisible between two labs. Can we both have our own workspace? Now the Companion is showing only one Workspace (xx Owner). Can I create a new workspace that is completely separate from the owner's workspace?

Thanks @marc ! Could you direct me to a link / documentation on how to do this? So far I have only found https://docs.pupil-labs.com/invisible/glasses-and-companion/companion-device/

@user-c2d375 I can create a new workspace, but how do I make sure that my recordings end up in my workspace?

Hi, we use the Pupil Invisible for a research project, and we cannot upload the data on Pupil Cloud because some data are sensitive. We would like to use the timestamps saved by the Companion App on the phone, we found the documentation on how they are encoded but we are struggling to decode and read them. Could you provide us with some tips or a procedure to decode these timestamps, preferably using Python ? Thank you in advance for your help !

Hi I am trying to change my work space but the device(phone) is taking a lot of time to sync. Plz help.

Hello! I was wondering, is there a way if i have a specific timestamp to know the exact time on a video ? For example the timestamp in ns : 1677583793620556032 what is the closest frame of the video that it's corresponds to ? Thanks in advance!

Good Morning! How I'm Able to check in what point of the video a specific gaze data referring to ?

Hello! After putting some markers on the surface of interest, I wonder if there is a way to obtain the surface coordinates during recording. I'm assuming using real-time api to get the gaze data and frame image, and processing the data on the host side, like marker mapper.

Hi @user-e93961! Technically this is possible, because the implementation of the surface tracker (that is what this feature is called in the Pupil Core software; its the same as the Marker Mapper algorithmically) is open-source. https://github.com/pupil-labs/surface-tracker

Practically speaking though, the implementation is currently more or less free of any documentation and thus very difficult to use. It's definitely on our roadmap to generate documentation for the repo, but its gonna take a while until we get around to it I am afraid.

If you are up for the challenge, I maybe have a couple additional code snippets that might be helpful. Let me know!

Hi @marc ! Is there a way to evaluate the gaze data (which parameters?) in different AOI? What I mean is % gaze, number of fixations, fixation duration, frequency of gaze shifts between AOIs etc. I have only the heatmaps in every single AOI but I need more data.

Hi @marc , thank you for the reply! I would like to have a try, could you please provide with the additional code snippets?

See e.g. the documentation of the Marker Mapper exports here: https://docs.pupil-labs.com/invisible/reference/export-formats/#marker-mapper

Hello developers, I'm Rohan. I've recently started using the Pupil Invisible for a test experiment, and I have an issue. The test goes as follows: I wore the glasses to look, and five individual dots, sequentially placed on a screen, are all separated by a good amount of distance.

Upon looking at the data/recording, it seems as if the gaze markers are off by a significant amount. Even though I'm looking at the rightmost dot on the screen, the gaze markers are stuck to the left and would only reach the second marker from the left at most.

I'm still an amateur, trying to learn how to calibrate, make things accurate, and ultimately learn. I would like to have any help from the community. Thanks

LSL Relay

Hi we used the Marker Mapper on many occasions and I’ve always placed the markers with the same orientation as they are given on Pupil page but do they have to be placed this way to work or does it not matter which side is on the top? We are just about to use them in a medical training research project where the clinicians will be setting up and so we want to make the process as simple as possible.

Thanks Eleonora that will make it so much easier for them to use. 👍🏻

Hi I am uploading the raw data exported to the pupil player but I am unable to see any video. there is nothing visible. I even tried reinstalling the player but i am getting the same issue. I am able to see the data on the cloud but not on the player. can I get some help.

yes i am downloading for pupil player

this is the error i am recieveing

@nmt This one?

Hello! Tell me, please, in the Pupil Invisible Monitor application, the sound from the microphone of the glasses should be broadcast?

Hello! I would like to stream my Invisible to my PC. Works fine at home via pi.local:8080 but not at work (academic institution). Could this be due to the fact that this connection is considered "not safe" (http instead of https)? Is there a workaround?

@nmt hey thanks for responding. I have done that multiple times and getting I am getting the same results.

@user-d407c1 I am using chrome

Sure I'll do that. Thank you.

Hello, I have a few questions, 1. After downloading the whole recording folder from the pupil cloud using the API, I noticed that the timeseries data is not included and I can't seem to find an appropriate API request to do so. 2. The recordings include files with .time format and .raw formatand I am not sure how to decode them. I would like to synchronize them and export all 3 cameras synchronized frame by frame labeled. I have seen the "Raw Data Exporter" from the pupil player that partly may solve this? But I would rather not have to use a GUI. https://github.com/pupil-labs/pupil/blob/master/pupil_src/shared_modules/raw_data_exporter.py

Hi, I forgot to change who the wearer was between my recordings and I have calibrated for each wearer at the beginning, is there a way to change the wearer afterwards so that the correct calibration compensation can be applied?

@user-d407c1 Hi, i have sent your the mail as per your request. kindly have a look and let me know how i can resolve the issue.

@user-d407c1 sry, I think attached the wrong one then. I'll send the pupil player format.

@user-d407c1 it working now. thanks😅

@user-60d698 Okay, got it! The timestamps included in the gaze.csv file coincide with the timestamps of the left eye video, i.e. there should be one row per video frame. For the right eye this is more complicated, as the timestamps are not included in this download. Instead you will indeed have to decode the .time files. The raw sensor files are not really meant to be consumed directly, so this is a bit inconvenient. The timestamps of the left eye camera video are in PI left v1 ps1.time and the ones for the right eye in PI right v1 ps1.time.

To read them in Python you could use

import numpy as np
np.fromfile(file_path, dtype="uint64")

Hello! Can't calibrate the Invisible. The red circle is not visible, so I can't to adjust the gaze. In the recordings errors are significant. Could you advice me how to make the recording more precize?

I followed your advice, log out and log in and as the result the App doesn't repy at all. What's the hell?! It is unreliable for work. I have the recording scheduled for today. What have I to do? to reject the project? or there is some possibiilty to come back to working regime?

Invisible app

Hello @nmt , @marc ! Our glasses get hot (about 55 °C). Please tell me what is the normal temperature for glasses to work for an hour?

Hello! Is it possible to calibrate the fixations in Pupil Player after making a recording?

Yes, I am using invisible. Can you point me to any existing code or a formula that I could use to make those corrections in the .csv file?

Hi! I conducted an experiment with the invisible glasses and want to upload it to the cloud. The experiment room had limited WIFI and thus the upload stayed at ~0%. Now, the "upload circle" is just empty. When I click on "upload" no visible changes are seen. It is also not uploaded on pupil cloud. Is there a trick to make the upload still work? I can see the whole recording on the companion app by clicking on it. It is just not uploading. I am also now in an environment with regular wifi strength

I should be in an environment with a stable connection now. Following the link gives me a "Site cant be reached; The connection was reset" error

Ah yes, that was indeed the issue. Thanks!

Hi, I wanna uninstall the Pupil Player v3.4.0 so as to successfully install the v3.5.0. But I came up with this error prompt all the time.Please help

Hey guys, I swear I saw a pupil guide for the invisible on running a study on mobile applications / for dynamic screens, does anyone know where I can find that?

Hi! I am trying to run the dense-pose-colab notebook. I uploaded the raw folder to my drive and ran the notebook. Unfortunately, I do not get a new folder that holds the dense-pose video and csv files. I renamed the raw folder from pupil cloud into something meaningful. Could this be the issue? I restarted the notebook and ran it again, but I still have the same "error". Of course I adapted the path and it seems to be correct.

My folder structure looks like this. Do I need to restructure it? I ran the notebook again but I still do not get a new folder in my google drive.

I get the run time error, which is, according to the comments, "normal", but other than that every cell gets a green tick. The issue I have is that I do not get the new results folder in drive, that holds the video and the two csv files.

On the last cell click on show code, there you will have the possibility to change the path of output. Please let me know if you still have issues.

While this indeed changed something (a % value indicating the current processing progress was printed ) it took ~4-5 hours to reach 76% and then the notebook crashed. Is this due to a poor network connection of mine, or is this still "normal"? Also a mp4 called denseposewas created. I cannot open it and it only holds 48 kbs.

Alright, I figured it out and it worked, thank you very much! One follow-up question: In your documentation, I saw that it makes sense to cut out portions where participants are looking somewhere else (e.g., their smartphone). Do you have a rule of thumb regarding how long of a 'distracted gaze' may be acceptable? For example, if we are talking about a 60-second clip of someone looking at a supermarket shelf and one second of that is spent looking at a smartphone or a person standing next to them, would you still recommend cutting out that part or would you say it doesn't matter too much since it makes up less than 2% of the overall video length? I hope my question makes sense. Thanks!

Hi there! I wonder if there is a easy way to get each frame in the scene video with corresponding timestamp.

@user-ace7a4 I've updated the repo and roll back the notebook. Although I was not able to replicate the problem.

Few things, please kindly make a copy of the notebook if you want to modify it, rather than disabling the test mode.

Second, would you mind trying it with a small section of your video using events? Please, create events on the cloud and download the raw enrichment again, which should include those events in the events.csv file. Then define them on the last cell as arguments using --start "event.name" and --end "event.name"

If you see any error, please let me know what the error states . If the error is a Drive timeout, you will need to not use Google Drive. Upload the folder on the side bar and change the input path and output path from Colab accordingly.

If that still doesn't work you can try substituting !python -m pip install 'git+https://github.com/pupil-labs/densepose-module.git'

on the 3rd cell with: `!python -m pip install [email removed]

which will use an older version of the pl-densepose module.

Also kindly note:

Google Colab notebooks have an idle timeout of 90 minutes and absolute timeout of 12 hours. This means, if user does not interact with his Google Colab notebook for more than 90 minutes, its instance is automatically terminated. Also, maximum lifetime of a Colab instance is 12 hours.

If you want to run it locally https://github.com/pupil-labs/densepose-module

Hi there! We discovered a problem, that after every recording we get the message "You have an unsaved recording from a previous session" and we can only start a new recording after closing and restarting the companion app. It occurs also after deleting all recordings. Is this a known problem or is there any solution to this? I hope you can help me.

(also, just to confirm that that isn't the culprit: I can work with the data transferred from the phone and exported via the player, right?)

@user-c9d495 unfortunately the change above means no override in Colab for now, I will try to fix it and update. But if you have a linux or mac, feel free to try it locally.

@user-d407c1 thanks for the update! Update from my side: I was able to run it and it is currently running locally (I noticed that while the colab cell didn't show outputs, downloading the nb and opening it in jupyter showed errors, whcih helped me fix it). So, as said, now it is running. However, your message suggests that next time I disconnect (which happens regulalry on colab, as you will know), it will no longer run with the --override? That's unfortunate. Is there a chance for me to change the line "!python -m pip install 'git+https://github.com/pupil-labs/densepose-module.git'" to some older version that would be the one I am currently executing (branch, flag?)??

Hello! is there a way (or a roadmap for) to fix gaze offset errors in the pupil cloud? I think we can do this offine (pupil player) but that means downloading all recordings and losing the event annotations we have made on the cloud interface. cheers!

Hi, new to using Pupil Labs invisible, but have run some successful trials with them last week. I'm struggling to get the app to sync, in particular the wearers are not syncing and it's not uploading to the cloud even though I'm connected to the internet. Still able to record footage, but not able to sync. Help appreciated!