Something else what I experienced and maybe it help you guys understand what is happening exactly. (Or maybe this is a new problem) Today I started to use the board calibration method because it is much easier and faster than the light method. I made a recording in the morning and I experienced a big difference between my cameras depth information. The difference is in space, not time! (After this I made 2 more recordings and they were perfect)
Is this the same issue or something else?

https://filetransfer.io/data-package/ul5e4Lr8

Any news regarding these issues? These seem pretty bad. Hoping to see a clear solution or explanation.
Cheers

Interesting news for my last issue. I checked again the wrong calibration file and when I run the board calibration it ended up being good. So I send my original and wrong iPiCalib file too, maybe it is help in the investigation.

Any news regarding the issues?

This looks familiar to me. In my case, I noticed a long time ago that this happens when iPi Mocap Studio gets confused about where I'm standing and flips the capture from one of the cameras.

The trick to preventing this error is to hold the board as far away from your body as you can--you don't want the data to get confused with your legs (like what's happening in your screen capture--this is most obvious in the right-side figure.) Also, hold the board in the middle of each side--avoid holding it anywhere near the corners. The tracker really needs to see the corners clearly for accurate calibration. From the tracker's point of view, the board needs to 'appear' to be floating on its own, so keep your body away from it.

If you're using a 180 setup, you need to hold the board with one had and stand to the side of the board. Hold it out with your arm straight.

Another very important thing to be aware of is how you move the board. Move the board with your body, not your arms--you don't want to pull the board near your body and confuse the tracker. Also, move board from side to side, not just front to back. (I also rotate the pitch and heading angle of the board but not too far because the board might bounce the IR rays away from the camera and hinder accuracy. I think I was told this angling motion doesn't matter too much, but I still do it anyway.)

From what I've been told, the back and forth motion is the most significant motion for accurate calibration, so take a few steps forward and backward if you can.

In short, proper board handling should make it very obvious to the tracker what it should be trying to align for each camera.

The material you use can also impact successful calibration. Make sure the board is rigid and has a matte surface. Poster board is NOT suitable because it's too floppy. Color is okay but avoid black because it can absorb the IR rays. White is okay if the room doesn't have really bright overhead lights or direct sunlight coming from the windows. Gray is probably best. FYI, I like to use a large sheet of white foam core board which you can pick up at any art supply store. It's very light weight so I can easily hold it up with one arm, and rigid enough for accurate tracking.

If you follow the above, board tracking should work perfectly every time. Good luck!

_________________
Greenlaw
Artist/Partner - Little Green Dog | Demo Reel (2017) | Demo Reel (2015) | Demo Reel (2013)


Watch a one minute excerpt on Vimeo now!

Duration of the captured data is also important. I try to make sure I have at least 20 seconds of usable data to track. Too little is useless and too much is a waste of processing time. Once you get the hang of it, you'll know by instinct if you've captured what you need.

_________________
Greenlaw
Artist/Partner - Little Green Dog | Demo Reel (2017) | Demo Reel (2015) | Demo Reel (2013)


Watch a one minute excerpt on Vimeo now!

Hello!

Thanks Greenlaw all of your help I appreciate that. All of your suggestions are great and I keep it in mind when I make my calibration file.
This calibration was "perfect" but when the algorithm finished the process and I started to work with it, the result was wrong. It is happened only twice, I don't know why. (When I run the calibration algorithm again on it, the file let it be good and usable)
The biggest problem is the merged video file desync. Any idea about it? Somebody experience the same?

Hello there!

Any news about the issue? I have some updates:

This issue is always there 95% of the time for me. I tried all of suggestions. Looks like this is not a performance issue. Both of my computers record at 30 frame/sec without frame drops. I use both of machines on high performace and I don't plug in any kind of usb device in the usb ports just the kinects. These machines are more then powerful for this task.
Almost every time the automatic merged video is bad. I have a big delay between the depth datas BUT when I merge the two video manualy with the video builder the new merged result is good and usable. I don't do anything special I just set the first and second camera records to 1 and 1 frame start.

I link a video (on the left side is the automatic merged video on the right side is the manual merge result)
https://filetransfer.io/data-package/hqJNoUwZ

Something is very wrong with the automatic merging algorithm.

We get exactly the same result with your recordings. That's because there is no actual time shift between videos from both PCs. But for some reason time synchronization between PCs on the start of recording produces a shift for several frames. We'll look on this issue closer this week. Hopefully we'll come out with a fix.

I wonder if you try to connect PCs via Wi-Fi network, will the issue persist in that case?

Hi @Zatesz
Please check out the new v4.0.6 of iPi Recorder. We've implemented some fixes in time synchronization during distributed recording. Hopefully this mitigates your desync issues. Let me know how it comes out.