invisible


Year

user-c6717a 01 May, 2020, 02:52:00

One more question, is there a way to enable OTG at all times? It seems to turn off after 10 minutes when not in use. Seems like something that could be easy to miss despite the warning. Thanks!

marc 01 May, 2020, 16:39:45

Hey @user-c6717a! Unfortunately OTG can not be enabled all the time. This is a limitation in OnePlus' versions of Android we can not disable. It is also impossible to have the app enable OTG programmatically. It has to be enabled manually by the user. I agree that this creates a risk of forgetting this when you start a recording. We tried to mitigate this as much as possible with the OTG warning and the sensor connection visualisations.

I still can not give you an estimated release date for the white paper. We are actively working on it with priority, but have partially been slowed down due to the COVID-19 outbreak. It will still take at least several weeks for sure.

Regarding your recording, it sounds like it was corrupted, which can typically only happen due to a faulty finish of the recording process, e.g. because the app crashed or the phone ran out of battery. Do you remember if you successfully finished the recording by hitting the red squared stop button, followed by "save"? Could you share the recording with us at [email removed] such that we can take a closer look at what went wrong here?

user-c6717a 01 May, 2020, 17:02:56

Thank you @marc! For the recording, It might have been that the battery went below 5%, but I know the phone didn’t die, so I thought I had saved it after the file stopped automatically. I thought that feature was supposed to allow you to properly save the file? Anyways, I was testing it with the battery fairly low (excited to try it out!), so I’ll keep an eye out. We will be using this in an experiment, so just want to understand all of the failure modes 🙂 I know the file is fairly big. Will those go through to that data email? Is there a particular file I would share?

marc 01 May, 2020, 17:10:02

@user-c6717a If the phone's battery drops below 5%, it will stop the recording and an LED in the glasses starts flashing within the field of view of the wearer. You would still have to confirm saving the recording to finish the process ( @user-0f7b55 @user-8f5305 right?). Could you zip the folder and upload that to some cloud drive, to which you can send us a link?

user-c6717a 01 May, 2020, 17:21:00

Yeah I think I saw the red light flashing on that recording and I think I hit save, but not 100% sure. I know I didn't hit discard. I'll upload the data a bit later today. Thanks!

user-c6717a 01 May, 2020, 17:25:24

Also, is there a way to adjust the calibration of the fixation after the recording is done? It looks like it was off to the right much of the time during a recent recording (although it seemed to become more accurate after I readjusted the glasses). I love the functionality being able to adjust that before hand, but it would be good to be able to fix it after the fact if we tell the participant to look at specific places during specific points in an experiment. It might even be helpful to be able to adjust for shorter segments of a recording where it seems off from a 'calibration' point. So far, I'm really impressed by what y'all have created with the Pupil Invisible. It seems to completely solve my problem of eye tracking outside in the bright California sun!!!

marc 04 May, 2020, 09:24:43

@user-c6717a Very happy to hear that you have success with PI in the California sun! 🙂 For some subjects it can happen that you get a noticeable constant offset in the prediction. The offset correction was added to tackle this problem. In theory this offset should be mostly constant across different slippage states of the glasses and lighting conditions, but there is some additional error for sure. Thank you for the feedback regarding this! I agree that it would be super helpful to be able to do this post-recording as well! We plan to add this as a feature to Pupil Cloud, such that you can add offset corrections to recordings online. There are a few other analysis features that take priority over this, but this will come in the not too distant future!

user-c6717a 04 May, 2020, 20:07:58

Awesome. Thanks and will keep you updated on our study's progress!

user-bea039 05 May, 2020, 09:10:44

Hi, I'm trying to detect surfaces using Surface Tracking Plugin. Could anyone tell me which size is the best for detect markers? I use 3x3 cm as a screenshot, but it is not good for detect markers. Sometime the plugin can detect, sometime cannot.

user-bea039 05 May, 2020, 09:11:58

Chat image

user-bea039 05 May, 2020, 09:12:11

Chat image

user-bea039 05 May, 2020, 09:13:26

I'm using lates pupil player (1.23.0) 😦

user-c5fb8b 05 May, 2020, 09:13:28

Hi @user-bea039, the markers seem indeed a bit too small. The best size depends on many factors, such as how far away you are from them, what resolution you are recording in and if you are moving your head (thereby introducing movement blur). We recommend you try different sizes to find a size that works stable for your setup.

user-bea039 05 May, 2020, 09:15:53

@user-c5fb8b Thank you for your quick reply! I got it, I try bigger markers.

user-c5fb8b 05 May, 2020, 09:16:17

@user-bea039 also you might get better detection by including a larger white boundary around the marker.

user-c5fb8b 05 May, 2020, 09:17:12

@user-bea039 did you use the marker sheets from our website?

user-c5fb8b 05 May, 2020, 09:17:51

to me it appears the white boundary is smaller in your setup that on our example sheets

user-bea039 05 May, 2020, 09:18:32

@user-c5fb8b Yes, I downloaded Apriltags tag36h11 0-23 marker sheets and print it out A4 paper.

user-bea039 05 May, 2020, 09:20:29

I use this marker sheet. https://docs.pupil-labs.com/assets/img/apriltags_tag36h11_0-23.37196546.jpg

user-c5fb8b 05 May, 2020, 09:21:26

@user-bea039 ok this should have enough white space around the markers. Be aware not to cut away the white space when cutting out the markers. Otherwise you just have to try a bigger size 🙂

user-bea039 05 May, 2020, 09:22:33

@user-c5fb8b OK, I make another makers with white space. Thanks a lot!

user-6952ca 05 May, 2020, 13:43:35

Hello, everyone! I have been greatly enjoying my experience using Pupil Invisible glasses so far! I was wondering if there is a way to extend the battery life of the companion device using some sort of external battery pack that would allow the glasses to also be connected. The phone does charge quickly, however, so extended recording sessions are still easy to do!

marc 05 May, 2020, 13:59:51

Hi @user-6952ca ! I am glad to hear you had a good experience so far! It is unfortunately not possible to connect an external battery pack. What you can do is to hot swap the Companion device to one with a full battery. This requires a brief interruption, but since there is no setup it would just take a couple seconds.

user-6952ca 05 May, 2020, 14:59:55

@marc Gotcha, thanks! Also, is it possible to grab recorded data from the phone directly without having to upload it to Pupil Cloud?

marc 05 May, 2020, 15:03:35

@user-6952ca Yes, you can download the recordings via USB to a computer similarly to how you would download e.g. regular photos or videos. It's the same data that you would download from the cloud. Let me know if you need more detailed instructions on how to do that!

user-c6717a 06 May, 2020, 03:49:29

Hi all, my new pupil invisible is no longer connecting to the the eye cameras and I can’t see an eye tracking fixation point. The world camera seems to be connecting. I’ve tried restarting the phone, uninstalling and reinstalling the app. Any ideas on how to fix this? Thanks!

wrp 06 May, 2020, 03:50:08

@user-c6717a did you by any chance update the OS on OnePlus6 device?

user-c6717a 06 May, 2020, 03:50:29

I did not update the device.

wrp 06 May, 2020, 03:51:11

OTG is enabled, correct?

user-c6717a 06 May, 2020, 03:51:30

OTG is enabled

wrp 06 May, 2020, 03:53:07

@user-c6717a could you please send an email to info@pupil-labs.com with your order id and screenshot of what you see in the homsecreen of Pupil Invisible Compaion App when you press the info button so we can see what sensors are being shown as connected? And we can continue this conversation via email.

user-c6717a 06 May, 2020, 03:55:42

Will do. Thank you for the quick reply!

wrp 06 May, 2020, 04:17:29

Thanks @user-c6717a email received. We will be back in touch later today.

user-bea039 07 May, 2020, 03:33:15

Hi, I tried to detect surfaces by Surface Tracking Plugin and finally I did it! Thanks a lot.

I have a question that is there any solution to fit a rectangle of surface to a real surface such as a display? An image from a camera of pupil invisible is distorted so defined surface cannot fit to the display.

user-bea039 07 May, 2020, 03:33:29

Chat image

user-bea039 07 May, 2020, 03:34:01

I want to define the computer's display as a surface.

user-c5fb8b 07 May, 2020, 05:00:57

Hi @user-bea039, what exactly do you want to achieve? Undistorting the whole image is an expensive task, so Pupil Player does not do this. Instead we e.g. take the distortion into account when applying the common operations for the surface tracker, such as mapping gaze to the surface or calculating heatmaps.

user-bea039 07 May, 2020, 05:31:32

@user-c5fb8b I want to calculate where a gaze position corresponds to a screen coordinate.

user-c5fb8b 07 May, 2020, 05:34:29

@user-bea039 the surface tracker will output normalized surface coordinates for the gaze points on the surface, i.e. x/y values between 0 and 1. When you multiply these with your screen size, you will get gaze data in pixel coordinates of your screen surface. Please have a look at the docs about the normalized surface coordinates: https://docs.pupil-labs.com/core/terminology/#surface-aoi-coordinate-system

user-c5fb8b 07 May, 2020, 05:36:24

The internal calculation of the normalized surface coordinates takes into account the screen distortion, even though the visualization of the surface (rectangle drawn on video) does not.

user-bea039 07 May, 2020, 05:38:06

@user-c5fb8b Oh, that's sounds great! I'll check it. Thank you!

user-627fd4 13 May, 2020, 11:41:08

@user-c5fb8b how do I retrieve the camera matrix of Scene Camera?

papr 13 May, 2020, 11:51:10

@user-627fd4 that depends from where you want to access it. You can start by copying our prerecorded intrinsics. Would that be sufficient for you?

user-627fd4 13 May, 2020, 11:52:19

@user-627fd4 that depends from where you want to access it. You can start by copying our prerecorded intrinsics. Would that be sufficient for you? @papr may I know where I can find the pre-recorded intrinsics?

user-627fd4 13 May, 2020, 12:11:13

thanks!

user-e8f45a 14 May, 2020, 14:45:29

I am in the process of selecting an eye-tracker for research purposes (gaze data on screens and real life situations). The invisible is probably the best suited platform. It doesn't require calibrating. But what are the gaze accuracy and presicion for this platform?

marc 14 May, 2020, 14:57:07

Hi @user-e8f45a! We are still in the process of writing a white paper that evaluates the gaze accuracy of Pupil Invisible in detail and I can't give you exact numbers yet. The accuracy is in the order of a couple of degrees. In lab conditions the accuracy of Pupil Core is certainly higher that the accuracy of Pupil Core. But in more challenging real-world condition Pupil Invisible quickly starts to outperform traditional eye trackers. For screen-based studies the conditions are often very controlled and the accuracy demands rather high (e.g. for reading studies). If Pupil Invisible is still applicable here depends on you specific application. For real-life situations it is well suited.

user-627fd4 15 May, 2020, 10:30:09

@papr is there a way that I can also fetch the camera position ?

user-627fd4 15 May, 2020, 10:31:08

The camera position in 3d space

user-627fd4 15 May, 2020, 10:31:53

I have certain ideas to do so but if pupil offers this feature directly, it might save time

marc 15 May, 2020, 10:33:51

@user-627fd4 We offer a feature like that based on markers placed in the environment within our analysis software Pupil Player. You can find documentation on it here: https://docs.pupil-labs.com/core/software/pupil-player/#head-pose-tracking

user-627fd4 15 May, 2020, 10:36:23

Thank you

user-e28285 20 May, 2020, 10:32:45

I am trying to build the Pupil Labs Invisible source code, but I am getting a GLEW error when running: 'python -m pip install -r requirements.txt' on Windows.

user-e28285 20 May, 2020, 10:33:13

The error that I am getting whilst building

Chat image

wrp 20 May, 2020, 10:44:45

Hi @user-e28285 I believe you're working on running Pupil Core from source. As there is nothing that one needs to build/compile for Pupil Invisible. Please repost your discussion in software-dev to get feedback.

user-e28285 20 May, 2020, 10:46:37

@wrp All right thank you. I am trying to build the Pupil Invisible monitor but after following the README I received this error, I shall repost this in the software-dev

wrp 20 May, 2020, 10:47:18

Thanks for the clarification @user-e28285 best to continue this discussion in software-dev

user-bea039 20 May, 2020, 13:33:39

Hi, I have a request for pupil invisible companion app that auto save feature when pupil invisible is disconnected. Movie and other files are lost when pupil invisible is disconnected before save button pressed :(. I think a form factor and wearability is very good and improvement of companion app makes pupil invisible much great! This is just a feedback, thank you 🙂

wrp 25 May, 2020, 06:45:47

Thank you for the feedback @user-bea039 - videos should be saved on disconnect during a recording.

However, the behavior you describe is a physical disconnect of the hardware (e.g. unplugging USB-C from Pupil Invisible Companion device) after pressing stop but before pressing save, is this correct?

user-bea039 29 May, 2020, 05:04:17

@wrp I mean before pressing stop and pressing save. I lost my recorded data accidentally I pull a cable that tight a phone and pupil invisible last month, before pressing stop and save button 😦

wrp 29 May, 2020, 05:10:08

Thanks for the feedback @user-bea039 - Pupil Invisible should be robust to disconnects - e.g. should continue recording even if the cable disconnects. I will discuss this with our team and try to reproduce the behavior internally.

user-bea039 29 May, 2020, 06:20:48

@wrp Thank you!

End of May archive