opentrack 2021.2.0

[sthalik]

Notes

Workaround Windows Defender false positive by compiling the release with Visual Studio 2022 (#1371).

Features

Add another slider to the PT point filter. Setting it to a low value prevents sawtooth-like motion. The point filter greatly reduces the need for using Accela or any other generic filter.

---

This discussion was created from the release opentrack 2021.2.0.

[Len62]

All is fine here with this version, no false positives. Have to play with the new PT point filter some tomorrow, thanks for the new update!

[bbecausereasonss]

I'm using Acuro could you please add the PT point filter? I find Acuro to be superior than most methods.

[Juanillus]

I have updated from 1.2 version and in my case the 2.0 version has a terrible jittering in DCS that makes it unplayable. Please, could you put a link to the 1.2 version to reinstall the olde version again?

[sthalik]

2021.1.3 is below. There weren't any changes that affect DCS though.

[Juanillus]

I tried 2021.1.3 also, and I had the same jittering issue with both 2021.2.0 and 2021.1.3. But 2021.1.2 worked fine.
I solved the jittering by downgrading to version 2021.1.0 that I found on other page. It could be some change from 2021.1.2 to 2021.1.3.
Thank for your response.

[sthalik]

That might be related to a change in the PT tracker. Basically, you need to adjust the threshold and min/max point size as the computed point size is different. So drag the slider a bit more to the right.

[Juanillus]

By the moment I'll stick to the older version, it works very well. Just if it can help your develope, in the newer version the extraction blue circles were shaking and slightly "ellipctic". Thank you.

[sthalik]

The points are the same for tracking purposes. It's only the visualization that's changed.

[3itZ]

I tried 2021.1.3 also, and I had the same jittering issue with both 2021.2.0 and 2021.1.3. But 2021.1.2 worked fine. I solved the jittering by downgrading to version 2021.1.0 that I found on other page. It could be some change from 2021.1.2 to 2021.1.3. Thank for your response.

Absolutely the same problem. I stopped at version 2021.1.2, which works correctly.

[sthalik]

Have you adjusted the target point size in the PT tracker with the slider already? It needs to be larger by a factor of 1.75 compared to 2021.1.2.

[3itZ]

Have you adjusted the target point size in the PT tracker with the slider already? It needs to be larger by a factor of 1.75 compared to 2021.1.2.

Before writing about my problem, I tried everything. Dot size, camera settings, point filter settings, etc. I fiddled with this for several evenings...

[WF000]

This is how i do it and is rock solid, well in BMS , no DCS yet.

[Len62]

Ok, i set-up using your configuration settings and it works nicely. I'll fire-up a game shortly for further testing and report back. Thanks WF000!

[Rayspit]

I tried to download Smoothtrack on my Android phone and it says that it's not compatible with my phone. I have Galaxy M12 phone. Thanks in advance

[Len62]

Ok, I fired up IL2 Great Battles and I must say using these settings has made movement much smoother and more fluid. It would seem that no filter with point filter checked makes quite a difference.

[WF000]

No prob. Well, 2cc , .. after a longer period I noticed a bit of jitter down -Z axis (zoom, translation forward/back).

Accela is still superior then Point filter as no any jitter, never... so I reverted to Accela...

One BUG for this version tho, Halik.

When output set as Freetrack , if selection is NOT on "Enable Both" (trackir and freetrack) , then it deletes the path to Naturalpoint reg entry.
So, no reg entry to ../modules/ if not set to both.
... of course TrackIR support is not working then.
I prefer freetrack disabled (only trackir) as no collision possible then.

Cheers

[dglava]

2021.2.0 is giving me really bad results. I am using the same settings (same profile, same filter, no PT filter) and the movement seems erratic and jumpy. It seems more sensitive and movement seems to involve unused axes as well (looking to the right moves it up an down). Have there been any changes which would explain that?

I tried the same settings as WF000, but with the same issues. On video it's subtle, but it's enough to make playing unpleasant.

Here's a video of 2021.2.0:
https://i.imgur.com/FkUel9d.mp4

Here's a video of 2.3.13:
https://i.imgur.com/CmAzQ0l.mp4

[sthalik]

@dglava confirm if this build helps. https://www.dropbox.com/s/6wre9q8wzawpp80/opentrack-test-20211214_01.7z

Otherwise I'm gonna have to revert more things.

[dglava]

Barely any change. I have a video of using it with the PT filter and without it if it helps:
https://i.imgur.com/kJlywLM.mp4

I have also tested version 2021.1.3 with the same results. Going one version back to 2.3.13 makes it behave in an acceptable way. I am not sure if the issue is due to my settings or tracking device though and I understand that it's not a good "bug report".

[Vortex76]

I have the same problem, jumpiness mostly noticable on yaw axis ( where movement is longer usually ) like its lost the smoothness it had on 2.3.12. I tried all settings proposed above and also the build you gave to dglava but it remains the same. Its like its tracking fine, but accella loses its charm, and at least for me I cant go further than 2.3.12, heck I just made an account here so I can write about it. Please help us. You rule!

[sthalik]

I need to know your setups.

1. Are you using the ps3eye open driver?
2. Does it help to switch to/from the open driver?
3. Are you using the point filter?
4. Is the jumpiness still present when using the Hamilton filter?

[Vortex76]

1. I use open driver a couple of months now.
2. No problems with or without it. A bit faster and more responsive when on it.
3. Yes and jumpiness on 2.0 - No and jumpiness on 2.0 - No and all good on 1.2.
4. Tried with Hamilton but I cant get it to stabilize. Is even worst than when on Accela or PF.
5. If it helps the jumpiness doesnt show in the octopus monitor. It is there in game ( DCS ).

[sthalik]

See if this https://github.com/opentrack/opentrack/releases/tag/opentrack-2021.1.2 version works. I believe that the problem was introduced in 2021.1.3.

[Vortex76]

Yes thats the last one that was butter smooth. I hope you nail it! We always have this one to roll back to, but we also appreciate your efforts to bring more goodies and new versions to us. Thanks in advance!

[sthalik]

So let's bisect this. I'm going to have to send you a couple of builds, but by the time we're done it's going to be evident as to which commit broke filtering.

https://www.dropbox.com/s/kxlfpcrvuhqr8ns/opentrack-bisect_01.tar.xz

[Vortex76]

05 Is jumpy and also movement on yaw outputs in pitch axis too as far as I can see. Gives out a very confusing outcome. Also has different feel with pf on and off.

[sthalik]

Good. So it's 76d9145 and indeed it's the point filter that's the problem.

[dglava]

I have tested the bisected builds and I'm getting the bad behaviour already on bisect_01.
2021.1.2 works as expected.

Are you using the ps3eye open driver?

I am using the ps3eye open driver.

Are you using the point filter?

With the filter it appears smoother, but it's still jumpy.

Is the jumpiness still present when using the Hamilton filter?

I don't have Hamilton available in bisect_01, but I am generally using Hamilton and experienced the jumpiness with the newer versions with it.

[sthalik]

@dglava can you message sthalik on irc.rizon.net, sthalik on skype, or join the discord at <https://discord.com/invite/NBVX6arwnv>?

[sthalik]

@dglava sent you a PM on Discord.

[Len62]

Sorry I can't join in on the testing this morning, will try later on today.

[sthalik]

@Vortex76 please confirm whether this helps, or at least changes something at all. https://www.dropbox.com/s/a66rrn2j7na7usx/opentrack-pf-test_01.tar.xz

[Vortex76]

Yes this is much better. Still not like 2.3.12 though.
Then, Pf behavior feels more logarithmic and rounded, but it introduces some lag so it makes me wanna have it off.

[sthalik]

I'm curious. What happens if you switch PT color mode to red only or green only? Is it any different than natural/average? I'm asking because if anything else than natural/avg is selected, then RGB mode is used with the ps3eye driver.

[Vortex76]

No difference with red green or blue. Feels same like nat/avg. If switching to any of the chroma keys it doesn't track at all.
But I m on Delan Clip which uses infrared leds so maybe others might notice different behavior.

[Len62]

@sthalik Ok, finally had a chance to look at 2021.2.0. Using accela filter operation is normal as far as lag, jittery movement. Turning accela off and using the new PF, operation is normal with the exception that movement tends to drift slightly, could be just a setting that needs adjusting in the PF settings. Using accela, if I turn my head I can stop at any point and everything stays steady, with the PF filter position tends to drift by a few degrees. I sometimes wonder if the jittery movement can be attributed to the games themselves and how new scenery gets drawn and compiled in memory. I haven't fired-up DCS recently, my experiences are with IL2 Great Battles.

[Len62]

Ok, tested in DCS World and with the accela filter enabled it was fine. When I switched to using the PF setting stuttering was more evident. Have to play around with PF settings to see if it any improvement is noticed.

[sthalik]

You need to lower Accela's deadzone to make the pf stutter go away. Normal pf settings are around 1.1 with limit set to 0.2 or slightly lower.

[Len62]

So, use accela and the PF at the same time? I was not selecting a filter when using the point filter.

[sthalik]

Yup. Use the low intensity point filter to stabilize the view for lower Accela smoothing, i.e. first slider. You can keep the PF setting at some 1.05, 0.2 and overall gain a faster response than with Accela alone.

[Len62]

Ok, here are the settings I used:


Things seem to be working really well in both IL2 and DCS. Will experiment some more over the next few days.

[animusVoxxx]

I don't like this new Point Filter in 2021.2.0_2 version. Want to use Opentrack with Accela only, as before but with new fps on camera.
PT makes view slow, with huge delay, i hate it. Ideal view for me is like human eyes must be fast and precise, easy to stop in place you need without wobbling. Tuned Accela do it well.

[Len62]

Ok, I'm still on 2021.2.0. Can't find the _2 version.

[animusVoxxx]

2021.2.0 is correct name. I have couple duplicate files in same folder.

[sthalik]

I disabled requesting grayscale from the ps3eye open driver. Does this build help you any? https://www.dropbox.com/s/6vhkrd2kmyfrpqw/opentrack-master-20220102_01.7z

And yes, Accela should function as it previously did without the point filter.

[animusVoxxx]

No changes for me with master version. Circles around diodes keep jittering with different settings.

[sthalik]

@animusVoxxx what are your PT thresholding settings? Try screenshotting the first PT tab.

[Vortex76]

Yep Thats it..!!! ( *opentrack-master-20220104_01.7z* ) Its pretty stable using old accella settings, with threshold 9.73, pf 0.13, limit 0.32 seems to work fine! I had to mess around a bit with the values to stabilise more but now its smooth as before the PF was introduced. One problem though is that for every different person will need that bit of fine tuning so still you ll have lots of people complaining..Maybe write a paragraph on what every setting does so its easier for the user to comprehend and adjust to his needs? Just saying.. Anyway, I think you nailed the problem. Thanks Stanislaw!!!!

[sthalik]

It should work just as well with Accela or Hamilton as before the point filter got introduced. Can you try out that configuration?

[Vortex76]

With Accella on and PF on is really fine as long as you tweak the settings. I use exaggerated values in Accela from before so I had to go easy on the PF settings. So :
Accela ON + PF OFF = OK
Accela ON + PF ON = OK
Do you need me try some specific setup?

[sthalik]

Something weird is going on. Can you notice how the cyan ellipse drifts off the point's center when using extreme orientation? For me it points to mostly empty space on high yaw.

[Vortex76]

try pf 0.13 and limit 032. That stabilised it a lot for me. The limit mostly

[Vortex76]

[sthalik]

@Vortex76 @animusVoxxx @dglava: Check this build: <https://www.dropbox.com/s/42f5v14amkmpz40/opentrack-master-20220105_01.7z>. I plan to release it at this state for 2022.1.0.

[3itZ]

20220105_02 slight twitching is present.
20220105_01 there is no such thing, everything is smooth

[sthalik]

I brought back the threshold and meanshift changes. This is the same as _01, but also has the freetrack fix. https://www.dropbox.com/s/fggp7823bluktbp/opentrack-master-20220105_04.7z

[Vortex76]

No problems for me with either 02 or 04, using the settings I posted above. They both seem to work really fine.

[3itZ]

20220105_04 seemed to work well, there was little time for the test...

[animusVoxxx]

20220105_04 is better, now playable without point firter. Stepping a bit.

[Len62]

No problems with 02 or 04. Tried Vortex76 settings, but found some slight stutter. Reverted back and feels really good.

[Perfectionman]

After closing OpenTrack I lose all settings and when I relaunch it everything is back to default settings, NOTHING saved. I have done uninstalls and made sure to delete all files and reinstalled. I have disabled OneDrive sync for the Documents directory. I have tried the portable version and an older versions but same result. I'm at a complete loss as for what to do :(

[sthalik]

I'm telling you, the sync hasn't been fully disabled. You probably have temporary files in Documents/opentrack-2.3 and that signifies that they couldn't be renamed. They're not present otherwise.

Anyway run the app through Windows Console by dragging the executable onto the window, then press Enter. You should get an error in the form of,

    qDebug() << "error with .ini file" << settings.fileName() << settings.status();