Eyes++ glitching

bonemap

I do not get anything like what you have demonstrated. So I am unable to confirm that behaviour. I did note intermittent, infrequent and very fast glitches but not enough to be able to capture as a screen grab.

https://recordit.co/7IPnnem9z7

best wishes

Russell

Woland

@fred said:

before I submit a bug report

I'd say go ahead and submit a bug report. We'll definitely need repro steps.

DusX

@fred

are you using a depth map as the input?

Fred

@dusx I see this with almost any input on any platform, I tested, Blackmagic capture on windows, depth map from Kinect 1 windows and Mac and webcam on Mac (as in the video)

barneybroomer

Did you test it with different values in the general service task?

DusX

@fred

I can also see a single frame flicker (not as pronounced as what you see) every 5-15 seconds.
I don't beleive its affecting tracking, since I added some visualizes and didn't see any value shifts that corrosponded with the monitor glitch.

I will add this to our bug tracker.

dbini

I saw some single frame flicker glitches during a workshop last week but I assumed they were more connected with my projector/signal cable and the fact that I was in an old school building in a thunderstorm. I wasn't using Eyes++ - I had a kinect1414 through OpenNI (Mac OSX High Sierra) - I didn't see it happen the following day in a different venue, so filed it away as an environmental issue, but this thread may indicate something more internal.

Fred

@dusx So I had a little more time to test, the glitch goes away when the system is under heavy load (it is running lower than the target frame rate - I am guessing this is due to a thread being unlocked and updated when it is not ready for a new frame - the higher system load must maybe makes it queue the frame).

Adjusting general service task has an effect, but does not make it go away totally.

Unfortunately when this glitch happens it is like cutting the video into the eyes++ actor and the blobs get new IDs, this really interrupts tracking. I can get closer to stability by adjusting the lifespan but obviously it is better if I can use that parameter without having to use it for this.

DusX

@fred

It does seem you are experiencing much more extreme glitches that I did. Mine were very short.. maybe a single frame, and I wasn't seeing any tracking issues.
That said its added to the bug tracker, and I am pushing it as a priority.

Fred

@dusx I get a frame or two, enough to re-identify the blob and interrupt tracking. Thanks for the priority, again it was for students who were confused that the IDs were changing.