Hello!

I bought an Azure Kinect about a month ago with ipisoft and it worked great, so I bought another so I can do some turns.

I went through the flashlight calibration which was perfect throughout the test, but for some reason I'm actually getting worse tracking than before. It appears as though my depth information is all artifacts, unlike before where the depth from a single sensor was perfect for the front angle. I'm also making sure my movements aren't sporadic, in fact they are actually very slow.

A few things I may be doing wrong:

- Darkening the cameras in-program while I do the calibration (should I be calibrating with this option off?)

- Maybe I'm not lining the floor with the program correctly?

- Our living room is very long, I'm curious if one camera is picking up too much depth (a lot of yellow for one side, but both cameras seem to be effected.)

I'd be happy to post pictures after my next test once I have the living room to myself again (working from home in a small apartment with my wife is difficult for capture.)

- Rusty

Tracking errors in multi-camera setup may be a result of poor calibration. And in some cases calibration can be actually poor despite it's reported as perfect/good - false positive from the quality detector. Do point clouds from both sensors align well in the scene?


Not sure what you mean


iPi Mocap Studio detects the floor automatically during the calibration process. However, it may go wrong if there are not anough floor points visible in cameras. Does point cloud look aligned with the ground plane in the scene?


Yellow (no depth data) should not be a problem unless it covers parts of an actor or much of the floor.


It would be better if you could share .iPiVideo files (via a file sharing service). You may email the links to our tech support address if you don't want to put them here in public.

Thank you for your quick reply!

I'm going to be running a few tests tonight to see if the calibration can be improved. If not I can supply you with the .iPiVideo that I get.