Hi I use the data collected by OnePlus 6. Can I use the player to get the original data of 200Hz or can I only get it through the cloud, thank you
Hi @user-4bc389! The 200 Hz data can only be computed in Pupil Cloud. If you download a recording from Pupil Cloud, you can open it in Pupil Player and the 200 Hz will be available. Otherwise only the real-time gaze is available, which for the OnePlus 6 is ~55 Hz.
Hi! I'm testing the Pupil Invisible in a sound booth with 24-loudspeaker array. Would the magnetometer or the IMU performance be affected by those loudspeakers? I've been using the MPU-9250 before and I've had the issue that it could not work perfectly surrounded by 24 loudspeakers. Second question, I want to measure both gaze and head movement in yaw corresponding to the loudspeaker locations. Can you please lead me to any publications or sources that worked on similar analysis? Thank you!!
Hello, we are trying to use the marker mapper to obtain fixation, gaze, and heat map data. However, our stimuli is presented on a slideshow.
Am I correct in assuming that we can only obtain one heat map per enrichment or per 3 markers? If we wanted to obtain several different heat maps, we would have to use different markers for every slide we want to obtain that data from, or is there a different way to setup it up?
Hi @user-2ecd13! Whilst you do need a new enrichment for each heatmap, you can use the same markers. If using the same markers, be sure to manually add events in the recording that correspond to when each slide was presented. That way, you can choose to run each enrichment on the correct section of the recording. Read more about sections here: https://docs.pupil-labs.com/cloud/enrichments/#enrichment-sections
Hi, my recordings have not been uploading to the Cloud today. I was able to do so a couple of days ago. Is the server down or something?
The server should be available. What is the behavior you see? Does the phone try to upload, but gets stuck in the spinning circle? What may help resolve this is to logout of the app (possible from the settings view) and to log back in.
Hi @user-455cca! The IMU we use does not contain a Magnetometer that might react to the electromagnetic field of other devices in the environment, so we don't see a reason why proximity to loudspeakers should have any effect on it. There were multiple projects implemented using the IMU to estimate head-pose during different interactions, but I am afraid I am not aware of publications that came out of those yet, that I could link to.
Thanks Marc!
Working with the Pupil Companion app, is there a way to load recordings (made previously with Invisible) from the Android file system back into Companion (for subsequent upload to Pupil Cloud)?
Hi @user-193e84 π. We have responded to you via email π
Thanks Neil!
Hi, we are testing Pupil Cloud for the first time today and ran into troubles after uploading to the cloud. We managed to upload the recording from the companion app, and can also download the recorded data from the cloud. However, after uploading Pupil Cloud states the upload is being processed which it has been doing for an hour now and we were wondering if this is the correct behavior? The duration of the recording is about 1min. Would be great if anyone could clarify or guide is in the right direction! Thanks in advance!
Hi @user-1b1acd! This is not expected behavior! For such a short recording even the worst case processing time should not be longer than a few minutes. It seems like we have an issue with our backend not accepting the processing jobs. Thank you for bringing this to our intention, we are looking into it right away!
Hiya Marc, just logged in because we are having the same issue in our lab currently, so it's great to see that it's not just us doing something wrong π We are super excited about using the new heatmap feature though!
Hi @marc - thanks for the fast reply! Sure thing - we are looking forward to test the system once the issue is resolved!
Hi @user-5a2110! Yeah, I think this is unfortunately affecting everyone. Sorry for the inconvenience! I'll let you know as soon as I get an update from the engineers working on the problem! I am happy to hear you are excited to test the new features and hope we have you up and running again in no time. I'd love to hear any feedback once you ran you tests π
@user-1b1acd @user-5a2110 Our engineers could fix the problem. There was a bug that prevented the processing jobs from getting started. Now the jobs are processing fine again, however a longish queue has stacked up now. Therefore it may take a while until all the job been completed. Sorry again for the inconvenience!
Thanks for the quick fix @marc - one out of three recordings has already been processed! Looking good so far π
Glad to hear it's resolving quickly! π
Hello! We are attempting to upload eyetracking data (each video is maximum 5 minutes) from the companion app to the cloud so we may download the videos for analysis. However, all our videos since September have not been uploading, they have been sitting at 0%. We have tried switching wifi networks, we have reset the phone. We are wondering if there is anything else to try to get our videos uploaded?
Hi @user-f0adbd! When you say you "reset the phone" do you mean you restarted/rebooted the phone, or did you do a full factory reset? Could you try logging out of the app (from the settings view) and logging back in, and see if that helps?
This is what we are seeing on the application
Hi @user-69e8af I'm responding to your message (https://discord.com/channels/285728493612957698/285728493612957698/930884288558297149) posted in the π core channel here since it relates to Pupil Invisible. Do you have automatic upload enabled in the Companion App? Otherwise please try clearing the cache: Log out -> Long press on the App icon on the devices home screen > Select App info > Storage and cache > Clear cache. Log back in again.
Hey, are the Pupil Invisible glasses CE approved?
Hi @user-331791! Yes, they are CE approved!
awesome thanks, is there some sort of documentation you could point me to? (couldn't find anything on the website)
Please see the document linked here, which contains all the individual norms and directives Pupil Invisible is complying with: https://discord.com/channels/285728493612957698/633564003846717444/897462722692808734
Hi - since there has been a problem with the pupil cloud server yesterday, I'd like to give a quick update. All the recordings from yesterday have been processed successfully. However, we uploaded a roughly 7min recording 4 hours ago and it is still processing. Could this still be the aftermath of yesterday's trouble (e.g., queued processing jobs) or something different? π
It seems like this is indeed part of the aftermath of yesterday. We implemented a quick workaround for yesterdays issue that did however slow down the entire infrastructure a lot. This is paired with a peak in traffic currently. The result is that your recordings are processing correctly, but the process takes very long to finish.
We are currently working on a proper fix for yesterdays issue that will also restore the original speed of the infrastructure. I hope that traffic will also normalize again tomorrow. Until then I am afraid processing jobs for new recordings will continue to take a long time unfortunately. π
I will look into it right away!
Hi @marc thanks for the fast update! No problem - fingers crossed that the issue can be resolved easily and everything will be up and running smoothly again soon. π
Hi! I'm trying to figure out how to use the new fixation detection for Invisible. In your release it says that "Fixations will be calculated automatically on upload to Pupil Cloud and be available in all exports", but I can't seem to find any fixation.csv files in my downloads. I tried downloading the files directly after upload in cloud, and I tried using the Raw Data Exporter - however in the latter, all I get in the download is a "info.json" file (which is very different from the export format mentioned in #Enrichments in your docs). I also tried exporting through Pupil Player but still without any information on fixations. How to I access the fixations? And why does my Raw data exporter not export any raw data?
Hi @user-695fcf! Using the Raw Data Exporter should be the right way to go. Was the info.json
files the only file in the enrichment download? In theory it should be one of several files as described in the #Enrichment docs. Could you let me know the enrichment ID of that raw data exporter enrichment (right-click the enrichment -> View Details), so we can take a look if anything went wrong there?
Hi @marc, the ID is 5a88e818-69c5-4f9e-ad29-76370c68c0a1
And yes, it is the only file in the download. I wasn't sure if the function only works on new recordings (from after the function was released), but this enrichment consists of an old and a new recording that I wanted to use to test it. I tried making a few enrichments on different recordings, but all with the same result.
Thanks for the ID, we'll take a look! You are right, fixations will only be available for recordings uploaded after the feature release. They can be added to older ones on demand after contacting us as well though. Either way, the enrichment download should contain more files! We'll investigate!
@user-695fcf What Browser did you use to make the download?
Also could you let me know the filesize of the ZIP file you downloaded please?
I just tried in Chrome - .zip is 7.3 MB and all of the files are there, including fixations! So maybe it's just a Safari problem.
Thank you! The folder is automatically unzipped on my Mac, so I don't know how to see the file size for that. The info.json file says 105,3 MB in the download menu (Safari) and 999 bytes in Finder.
Thanks! I was just about to propose using a different browser! The issue with Safari is that is has a setting where it automatically opens "safe" files after download. For ZIP files this means extracting them and only keeping contained files that are considered safe. For some reason CSV and MP4 files are not considered safe, and thus only the JSON files remains.
This setting can be turned off in Safaris settings, which leads to a proper download of ZIP files which can be manually extracted fully.
@user-695fcf Would you like us to add fixations to your older recordings as well?
Ahh I see, thank you so much for the help! I will turn it off in Safari so I won't run into this issue again. And if you could add fixations to the four recordings from 2021-11-24, that would be great π thanks!
Will do! Those recordings should appear as "processing" again briefly. Computation shouldn't take long.
Hi, is it possible to upload a sinlge recording from the companion app in stead of all the recordings at once. I can not upload all the recording due to privacy issues, but I also don't want to delete them all from the app already. It can also work if I can upload the exported data folder from my computer to the cloud, but I also did not find out if this is possibe.
Configuring the upload per recording is not possible, but there is a not officially supported workaround that might work for you. You could temporarily remove recording folders off of the phone to a computer via USB (without exporting them in the UI), upload remaining recordings, and move them back to the phone after.
If you want recordings that have been on your computer to be still potentially uploadable to Pupil Cloud, it is critical to not open the in Pupil Player in the meantime, because it would alter your recording formats and render them incompatible with Cloud. This approach may not work for future versions, but works fine currently.
Okay thank you!
Hi - we were running the marker mapping enrichment in Pupil Cloud for various sections (between corresponding event annotations). While the extracted fixations do correspond to the timings of the sections, the generated heatmaps do not.
We expected that the picture, that is used as the base for the heatmaps, would correspond to a video frame within the defined time period (e.g., the 1st or Xth frame). We also tried defining the marker surface at the specific time when a new section would begin.
However, the resulting heatmaps (their base picture) do not relate to the picture that could be seen during this specific section of the recording. Additionally, different enrichments (all relating to different sections of a recording) were using the same (wrong) base-picture for the resulting heatmapp.
Just in case you would like to take a look for yourself, here is an exemplary enrichment-id: 0f89742a-4d2d-47a0-9ca5-1d0ab83bdaa3
Unfortunately, we couldn't figure out what causes the problem or how the picture for the heatmap is chosen. Any ideas?
Hi @user-1b1acd! The background picture used for the heatmap visualization is determined when a surface is defined for the first time. The image will be cropped from the video frame visible when clicking "Define Surface". So if you create a new Marker Mapper enrichment make sure you have the desired video frame selected when clicking the button. Note that the same image will be used for all recordings the enrichment applies to.
Can you achieve the background images you desire this way?
Hi @marc - thanks for the explanation! What you described is exactly what we thought should work. Unfortunately we still couldn't achieve the required background images. Here is the process I just tried:
1) Create new template for testing 2) Create new recording and upload to Pupil Cloud 3) Create new demo project 4) Add uploaded recording to new demo project 5) Add new annotations to the uploaded recording (using new name-pattern) 6) Add 2 new marker mapping enrichments (using original names too) based on 2 distinct segments (as defined by the newly added annotations) of the new recording. Also, when clicking "Define Surface" the desired frame was selected. 7) Hit "play"-button for processing.
Oddly, the resulting heatmaps are still using the same background picture. Is there some obvious mistake in this process that I am not seeing?
@marc hi Marc, could you plz help me with transferring data from the device to my pc?! It seems that my pc doesnβt recognize it or it needs some apps or something! Tnx
Hi @user-b811bd! Sure, I'd be happy to help. Could you let me know which operating system you use on your computer?
Windows
Okay, then you simply need to allow file transfer to the computer on the phone. The following steps should do it:
Open the notification menu on the phone after connected to the computer, tap Android system -> charging this device via USB, expand it and select "File Transfer".
I do the all steps, but as I choose the transfer data option, it goes on no data transfer by itself!
@marc it seems that it doesnβt allow me
Huh, that is odd. Could you try rebooting the phone and using a different USB port on your computer?
And potentially restart the computer as well π€
@marc I did use diff cables, restarting my pc. Nothing worked.
Also restarting the phone?
Yup
Iβve had this problem for weeks! I canβt use my recordings unfortunately
Have your tried something like Android to transfer them over wifi?
What phone are you using and is it updated to the newest OS version?
What you mean by Android?
Sorry, Airdroid is what I meant
Itβs the companion device
Is it a OnePlus 6 or OnePlus 8 device?
8
I havenβt tried Airdroid
Hm, the issue must be somewhere within Android or Windows, so not in our direct zone of control. Let me consult with my colleagues to see what ideas they have.
Airdroid would allow you access your phones storage via a browser interface to transfer files over wifi. It should work as well for recording tansfer.
Got it, Iβll try it meanwhile youβre working on the issue
@user-b811bd Additional input I got from my colleagues: - You have already tried multiple USB cables, but are you sure one of them was made for data transfer? Many charging cables are for charging only and can not transfer data. The original orange charging cable from OnePlus, which you probably tried, can transfer data though. If your computer has a USB-C port, the Pupil Invisible connection cable should also work.
Iβll try them, thank you Marc. I hope these measures solve the problem, if not, Iβll reach out to you later
@marc hello again, I wanted to let you know I was able to transfer the files to my pc using a cable! Would you mind plz LMK how can I have the merged video from both the scene and eye cameras? I mean the scene video that has the red circle?! Thank you in advance
You can either use Pupil Player to make this export, or the Gaze Overlay Enrichment in Pupil Cloud.
Thanks, where can I find the link for downloading it plz?
@user-1b1acd Thank you for trying! I think I could reproduce the issue myself. Let me investigate this a bit!
Many thanks for investigating! π I look forward to learning what may have caused the problem.
You can download it here https://github.com/pupil-labs/pupil/releases/latest#user-downloads
Thank you
I am hoping to use the gaze_positions.csv
file (specifically the norm_pos_x
and norm_pos_y
columns) to produce some plots of specific frames in Matlab in which I aim to overly the gaze position using Matlab's plotting functionality. I am having trouble, however, syncing up the norm_pos_*
points with the gaze position cross-hair that we get from Pupil Player. Specifically, when I plot the points denoted by the x and y coordinates contained in the norm_pos_*
column, it doesn't appear to be the same point as the cross-hair (a magenta dot in our case) that exports with the world.mp4 video. Does anyone know what might be causing this discrepancy?
Below is an example of what I am observing. The magenta point exports with the world.mp4 video from pupil player, but the norm_pos_*
columns in gaze_positions.csv at that same frame appear to point to the blue dot
How do you extract the frames from the video and how do you do the matching to the corresponding gaze datum? Some video decoders are not frame accurate making it impossible to get the matching right.
@papr I extract the frames from the video using Matlab's VideoReader
function, which creates a VideoReader
object that allows you to iterate through each frame one by one. To associate the norm_pos_*
with each frame, I tried two methods. First I used the nearest world_index to each frame number. I later read on the Discord chat archive that this method might be unreliable. So I also tried extracting the norm_pos_*
data that are closest in time, based on the timestamp associated with each frame (contained in world_timestamps.csv
) and the timestamp associated with each row of gaze_positions.csv
. This seemed like an imperfect solution as well: for some frames the dots appeared aligned, in others not.
@user-1b1acd It turns out that the implemented behavior is not what I described above. The algorithm will instead choose the frame of the video that has the largest number of markers visible by itself. With the current implementation there is thus very little control over what image is chosen. I will discuss changing this behavior with the team to make it possible to control the background image, but a release of such a change would likely take a few weeks.
Thanks for clarifying! That's very unfortunate - could you maybe give us a heads up once the issue has been discussed? It would be great to know if and when this feature will be updated, since background image control is a key element for us.
Can anyone advise if updating to OnePlus 6 OxygenOS 11.1.2.2 is okay?
No, this update is not okay to install. OnePlus 6 devices should stay on Android 8 or 9.
Thanks for confirming!
For sure! I am pretty sure we can address this in some way in the next release, which should be in 2-3 weeks. I'll let you know once I know more!
Thanks @marc , sounds great!
@marc hey Marc. I downloaded and installed the pupil player, but it doesnβt allow me to drag the files into it! Any idea?!π
Hey, what does the grey window say onto which you dragged the recording folder?
@papr When I open the player, first one black window appears, and then a bright (grey?) one. It says drop and drag the files. But when I want to drag them, it doesnβt allow
The text should change to reflect the reason for the failure. Could you share the error message with us?
I donβt get any error message. When I want to drag the file as soon as I put my cursor into the window one red circle with one line on it appears, which means I canβt drag
OK, thanks for the clarification. I have not seen this issue before. Can you confirm that you are dragging the recording folder, not a single file?
Yes, I tried both, the folder and the files
Which version of windows are you using?
Windows 11
Has that been released publicly already? We have not tested the software with this version of windows yet. For the window management incl drag and drop, we rely on a third-party library. I will check if they are aware of this issue already.
I also tried on windows 10 in another laptop but same issue
@papr do you see that black circle?
I do and thank you for checking. Am I seeing it correctly that you are trying to open the recording directly from the phone? Please copy the recordings to your computer first, create a backup and then open them in Player. Player will modify the format inplace to open it.
There are so many details with it
Yaaayyyy! Problem solved
Thank you again
I appreciate ur time
Thank you for giving it a try on Windows 10 and sharing the screenshot. Without it, I might have had to install Windows 11 and come to the frustrating result that it would have worked for me but not for you. π
It is actually working on windows 11 as wellπ
Please let us know if you notice any other issues on Win11!
Sure
Hi! We are currently evaluating the use of pupil invisible eye-tracker during visual rehabilitation of people with a visual field defect. The recordings are used to give feedback on their scanning behaviour. However, the size of the recordings is often very large and due to privacy it is not possible to upload the data to the cloud. It takes therefore quite some time before the recordings are uploaded to the computer to watch the recording together on a big screen. Time that is not available in one training session. Is it maybe possible to also stream the recordings from the phone to the computer as is done by the invisible monitor for the real-time video feed? If not, is this something that pupil labs might work on for the future?
Hi Eva! It is great to hear you are using Pupil Invisible in a visual rehabilitation clinical setting. In response to your question about real-time saving of recordings for immediate preview, there are a number of options:
1. Post-hoc Screen Stream If all you require is immediate playback of gaze overlaid on the scene video after the recording session, you can use this app to stream data from the phone onto a big screen: https://play.google.com/store/apps/details?id=info.dvkr.screenstream&hl=en&gl=US
2. Screen Recording: You could make a screen recording on the connected PC of the video streamed by Pupil Invisible monitor during the recording. Then, you would not have access to the raw data files immediately, but could review gaze data on the scene video straight away. Note that there is latency associated.
3. Network API: You could write some custom code using our Network API for Pupil Invisible: https://docs.pupil-labs.com/developer/invisible/#network-api It would then be possible to save data real-time. However, the quality would not be comparable because of the latency introduced by streaming over the network, and this would require some technical expertise.
I hope this helps!
Thank you Richard. I will look into your suggestions a bit more and see what works for us.
Hi, I have two One Plus 6 companion devices that are nearing their storage limit, is there anywhere I can find best practice info on clearing previous recordings in one go? Without removing them from pupil cloud of course? Many Thanks
Hi! Deleting a recording from the phone will not delete it in Cloud. If a recording is showing that it has uploaded successfully, you can safely delete it. Recordings in Cloud can only be deleted via the Cloud UI.
Thank you!