I would like to know how long is the delay between sending a command and the start of recording when controlling Pupil Invisible via API ? Assuming that there is a time difference of about 1 second between the clocks of the computer that sends the command and the cell phone that receives the command, is the timestamp recorded in 'events.csv' downloaded from Pupil Cloud using the time of the cell phone or the time of the computer that sends the command?

We are having trouble calibrating/finding instructions on calibrating the invisible PupilLab glasses. Can anyone help? 📣

Hi there,

I'm confused... I’m trying to write a program allowing to impose an image on the worldcam recording of the invisble or neon (post processing) that depends on the gaze. I’ve used the snippit code provided to read the raw data. Somehow I experience large differences when comparing the snippit using gaze ps1.raw and the csv coordinates provided after pupil player export. I was hoping to use the raw data since it does not a timely export. I thought it used to work fine, but now it does not... Or is there any way where you can run the command using python to obtain the gaze and imu file?

Hi @user-e91538 👋. The Invisible system doesn't have a calibration procedure in the conventional sense, but it does have an offset correction feature. You can read more about that, and how to use it, in this section of the docs

Thanks! I'll look into it

Hello. I have conducted a study with pupil invisible glasses. The participants were sitting in a simulator and completing a excavation simulated task. I have 3 AOIs in my experiment and when I export the enrichment data I notice that a lot of data is missing. This is an example with raw fixation data merged with the 3 AOI data in one file (n/a are missing values) The missing data also does not match in between AOIs, meaning that one data referring to a fixation is present while another is missing completely.

To be noted is that I am using the marker mapper enrichment to define AOI

Is it possible to have a better data export?

HI @user-5fb5a4 ! Would you mind sharing whether you created the enrichment and then added recording to that project? Also could you share the enrichment ID such that we can further investigate?

Hello. I created the enrichments after adding the recordings

40f0b276-d7ce-45a6-85f3-c7f7a87dd5ad

62f70e95-e5fa-41ab-905e-b0cc12f63aab

d3abd7dc-4ada-4f47-b1c7-102c86f6a13d

Hello, I have a question about cloud upload. Is there any expected delay between upload time and the data shows up in cloud? Today, I uploaded data but can't see it yet in cloud.

@user-480f4c Yes, Yes, and No. Interestingly, Everything looks fine, and videos uploaded in 2023-12-25_14:23:32 looks just fine.

@user-480f4c , It looks properly uploaded in the app, but not see in the cloud.

Hello! Using the Invisible model for wayfinding and signage research, I often find that the gaze point will disappear for a second or two throughout the video recordings, anyway we can avoid this from happening?

Hello Miguel! thank you for your feedback, it is more clear now. I also noticed that sometimes the detection of the fixation overlap even if the surfaces don't (e.g. I get 'true' in screen and wheel ). I guess that is a matter of precision and how close the surfaces I designed are. Do you think those can be improved?

Hi! I am running experiments with four pairs of Pupil Invisible glasses and I have managed to remote control one pair of glasses with the Pupil Invisible Monitor. I was wondering if it is somehow possible to remote control all pairs of glasses with the same computer?

@user-2251c4 kindly note, I have edited my response, to be more precise

Hi all, I am using Pupil Invisible, is it possible to do Apriltags detection offline locally? I know currently the process is to upload the data to PupilCloud but I sometimes need to do some post corrections on the gaze data and for Pupil Cloud I cannot do it so am exploring ways to do correction offline on raw data and then run apriltag detection to map the data to the relative coordinates to the surface.

We used the Pupils Invisible in an urban setting; having participants standing for several minutes at different public squares; now we’d like to analyze were the participants have been looking at during the time standing in the public space (cars; pedestrians; buildings; green areas; etc.). I am wondering what the easiest approach to gather these data would be and would be very happy to get some insights on that.

Hi! I'm working on combining the Invisible with a position tracker to calculate a full eye-gaze vector in 3D space. For this, I would need the yaw & pitch rotations for the eyes, idealy for each eye but combined is also okay. How do you suggest I go about this? Can I reliably convert the gaze x & y coordinates I get though the streaming API into a yaw & pitch, or are there better ways?

Hi, i am trying to create a programm by which i start recording by "space" key and stop it by "x" key. Starting the recording works well, but stopping doesnt. Whenever i press "x" the pygame window is not responding anymore, i can't exit the programm without restarting the Kernel and the recording on invisible companion also isn't stopped.

The only problem VS Code shows to me is: Import "pupil_labs.realtime_api.simple" could not be resolved and Import "psychopy" could not be resolved

I have no problem connecting to the device though to start, just the opposite to stop.

Can send the code too, if that helps

i noticed i generally have the problem that after 4 seconds the recording was started the program is stuck on "not responding" and any key i press kills the program so i have to restart the kernel to end it

Hi @user-5ab4f5 ! Without having a look at the code is a bit hard to provide more specific feedback but did you check the async version of our API? It allows for non-blocking operations, enabling concurrent execution, while non-async code executes sequentially and can block on IO-bound or long-running operations.

Also note that to start & stop recordings you could simply do POST request to the endpoint.

like this:

import requests

url = "YOUR_IP:8080/api/recording:start"
response = requests.post(url)

This was basically my code. @user-d407c1 Do you see anything off about it? Otherwhise i'll just try to reqrite it with the information you have given me and hope it works out

Hi, When I tried to run the script on my home wifi, the device connected and worked as it should. But when I try to connect the wifi at my friend's house (Maayan) or to a hot-spot from my phone - it doesn't work and there is no connection to the device. On the other hand, when Maayan connects to the device from the wifi in her house, it does work for her.

** It is important to note that I made sure that the computer (which I ran the script) and the mobile device were on the same wifi.

This is the relevant part of my code:
# Look for devices. Returns as soon as it has found the first device.
    print("Looking for the next best device...")
    device = discover_one_device(max_search_duration_seconds=15)
    if device is None:
        print("No device found.")
        raise SystemExit(-1)

The output: No device found. 

Hello! Is there a field or article where I can learn in detail about pupillary data concepts, such as azimuth angle? I am trying to better understand exactly which data these concepts correspond to. If you have any information on this and can help, I would be very happy.

Hi @user-228c95 ! There is no pupillary data with Pupil Invisible, would you mind confirming that this is the eye tracker that you are using? Regarding azimuth and elevation, you can find some information here

Hello. When downloading the raw data from the cloud is there also a file with saccade data? I see fixation, gaze, and blinks data but not saccades

hello, thank you for the response. What would happen in the case there are actually smooth pursuit movements? I guess that they are just counted as fixations by the algorithm

Is someone from the sales team available ?

Hi @user-c20bf6 ! You can direct your sales queries to sales@pupil-labs.com

Thanks

Hi, can you send an example of the proper way to insert the corners through the CMD?

HI @user-d407c1 We only see the gaze point disappear in the Cloud (it remains constant when viewing through the app on the phone. We see it disappear even when the Blinks are turned off when viewing on Pupil Cloud - any idea what it might be?

Hi, we've been thinking of using the invisible for a simulation drive study, but have strict restrictions about cloud data. Is there a way to directly integrate data from the capture in iMotions and then export it with the other sensors? As opposed to post-importing it to iMotions and linking it to the sensors? We currently have the invisible set available, but are open to other models if it's better. Thanks for the info!

Hello, I am event-tagging some of my videos recorded with Invisible on Pupil Cloud, but when I seek the video using the playhead, I keep getting this error message, and the video stops buffering, I have to refresh the page, after which it works for two or three 'seeks', before the error message reappears again, and the whole cycle repeats. As a result it's impossible to do the tagging efficiently.

I can confirm that this is only happening to one particular video in my project - other videos don't seem to have this problem. Can anyone give some pointers on how this can be overcome?

Hi there. I am using the real-time API. I am trying to get the eyes video frame using device.receive_eyes_video_frame(). While I can get the scene video frame and gaze data, this does not return anything. Do you know why this could be?

HI @user-00729e ! Eye videos streaming through the realtime API is only supported for Neon.

In Pupil Invisible, the eye cameras are positioned on the side and they are normally of not much use other than for our network. Since the view of the eyes is quite lateral it is hard to get meaningful information from that view, and for this reason, the streaming of eye videos for Pupil Invisible was never migrated to the new realtime API.

That said, if you want to obtain eye images (although they might not be of much use as I mentioned), you can still use the legacy API

They should be as described there "PI left v1" and "PI right v1". When running the snippet provided all sensors available should be shown in SENSORS

when trying to create a reference image - what does "form not ready" mean regarding the recording?

Hi @user-91d7b2 ! This usually means that you tried clicking on run but either the name of the enrichment, or any of the required fields (scanning recording, reference image,...) was not ready. Were you able to create it? Or do you need assistance with ?

Hi, I'm currently trying to use "pl-dynamic-rim" on one of my videos and I get the error: No valid gaze data in RIM gaze data for recording ID 39cf919f-1785-403c-9b04-371a79ed2dbb

This issue seems to persist with another video as well, resulting in two corrupted videos out of a total of five. I have attached the gaze data (reference image) file for your reference.

In the recording it seems that I did look at the screen, how can I check if the reference image is found? I can get the heatmap of it

There is a point when I gaze at it, and I did look at the image at least once in the video... So how can I fix it? or avoid it in the next recordings?

Hi  @user-91d7b2 ! Could please indicate the steps you follow? eg. Create Enrichment - Reference Image Mapper - Place a name for the enrichment - Select scanning recording - Upload reference image and click Run ?

Can you confirm it started after you clicked on run? Alternatively, you can invite me to your workspace to have a look.

Thank you, I'll check that! I have another question - is it ok to take a video with one eye closed or will it create problems with the data of the video?

Hi, There is a problem using reference image mapper, it always shows 'running' and I have waited for more than 20 min, the video is very short (30 seconds). Could you tell me what's wrong?

In the end, it reports there is an error. Could you tell me how to fix it?

Thanks for replying. I tried to make high quality recordings and did the reference image mapper more than 10 times, it still fails. Another thing is the new instruction in the new version really confuse me. In the old version, firstly, we should click a video(file1) recorded by glasses and in which record your gaze, secondly we upload reference image(file2) and scanning recording (file3) to create a 3d model , then we can map the gaze form video(file1) to your image(file2). but in the new version, there is no step to choose video(file1). maybe my understanding is wrong?

Yes i will invite you to the workplace. How do i do that?

Hey everyone. I am new to eye-tracking and Pupil labs. I have an idea for an upcoming thesis topic. We have the Pupil Invisible glasses in our lab. I would like to extract the pupil diameter as part of my data. However, as per the documentation, Invisible does not collect PD data directly. Does anyone know a way where it can be obtained from a video recording?

Hi, I am also new user of invisible, and would like to ask about the meaning of each word of fixations excel file.

What is the meaning of "azimuth[deg]" ,"elevation[deg]" ?

If there is any instruction documents which mentioning aobut this, please let me know.

Thank you Dom. I appreciate it. Then, these degrees are from observer's eye point, not from previous fixation to next fixation, am I right?

Thank you @user-4c21e5 . Would you suggest to shift to Pupil Core for pupillometry then?

Hi @user-4a6a05 , Could you please tell me which adapter I should use to be able to charge the OnePlus while I am measuring the gaze behavior with Invisible? I would nee d a Usb-c to 2 Usb-C adapter I guess, but what about the transfer rate? Thanks in advance.

Hi @user-cc9ad6! the scanning recording is a normal recording made with the glasses, the difference is that you do not need to be wearing them, it has to be under 3 min and it is not used for the metrics. You can read more here

I have invited you to my Workspace. And my problem is that: Now I want to map gaze from a video (which the view just move a little) onto a static image. According to the instruction, the scanning recording need to scan the view from different distance and angles. and it seems the Could only map the gaze on image from the scanning recording, instead of normal videos recorded by the glasses, right? By the way, I use pupil invisible, not Neon. does it matter?

Hi Nadia, Thanks a lot for your work. But I don't know how to select the gaze video instead of scanning video on the toggle above the preview when looking at the enrichment output. You mean in the visualization? In visualization, it shows we can select video but I couldn't find a button to that (see the pictures). Another thing is could you help me look at the 'robot2' project? I really want to map gaze from gaze video onto image by using a image and scan recording. But it seems I couldn't.

But as you can see, differ to the scan recording, in the gaze video, the view just moves a little, instead of including all distance and angles, that's why I always failed before. Does that mean I can't use this mapper in this kind of video?

Yes, I do that according to the instruction. the 'scan recording' is made by holding eye tracker glasses whereas 'gaze video' is made by wearing the eye tracker glasses

Thanks a lot! Now it is quite clear. I am very appreciated for your solution and patient explanation!!! That's so helpful for me.

Hi, I'm trying to use the enrichments,. I uploaded the reference image, the scanning recording, I chose the events where I want the enrichment but after minutes of waiting the system gives me an error....I can't understand where I'm wrong

....this is the situation

Hi @user-e91538! Would you mind sharing the details of the error? Based on the image you shared, it seems that the enrichment is still running, but I can't see any errors

no particular message....the little wheel on the top left keeps on spinning around and after many minutes (maybe 15) it stops and I see the word Error

@user-e91538 thanks for sharing more details about the error. This error is likely the consequence of a poor scanning recording and/or a suboptimal reference image. A good scanning recording needs to record the object of interest (in this case the painting) from all possible angles and from all distances a subject may view it. I highly recommend checking our scanning recording best practices. https://docs.pupil-labs.com/neon/pupil-cloud/enrichments/reference-image-mapper/#scanning-best-practices

thanks for your reply and suggestion.....I'll try with new scanning recordings

Hi there, Again the same problem, but now using pupil cloud. When i look at one of the movies, the movie is mirrored (you can for instance see this from the letters of signs) I'm we can also download the pupil player format. But still are struggling with the alignment of the pupil download and the pupilplayer format. I don't know where to start. Making a scatter plot between both x-axes provides a cloud Is the first sample different (pupilplayer format provides x and y coordinate, wheres cloud seems to start after a few seconds)

Hope you can shed some light

When looking again at the source, it is not mirrored. As said I'm confused... Maybe I'm just mistaking.

I'm trying to calculate it to the worldcam times... For the downloaded format from the cloud I used the time stamps of the worldcam

For the pupil player format I'm using the index of the worldcam.

I'm taking the average

I'm trying to place an image over the worldvideo. The location of the image is gaze dependent. In order to do so, as a first step, I'm trying to determine if the gaze as calculated from the excel is identical to what I see happening in Cloud/pupilplayer. Somewhere I'm making an error

Hi Nadia, Could you tell me how to get the gaze position on the image after using Reference Image mapper. Now I can only get the gaze position on original video, not image.