Hi guys,

after struggling around with the automatic calibration process - in my case for 3 Kinect setup - and most of the time getting a position error above 0.050m I wondered if this makes sense.

The calibration file comes as handy xml-filetype. Based on an already existing calibration file (error above 0.050m) I did some fine adjustment within the xml, saved it and immediately reloaded the edited file into Mocap Studio. With the help of the initial T-Pose I was able to check if the adjustment was correct, step by step. Also I could adjust every camera on its own.

The Question is, does this approach only work to get a perfect alignment of the T-Pose but getting (position/rotation) errors on the continuing mocap run?
Or is that the way you guys at iPi Soft would do as well to improve the calibration?

Also, is there a way to check what position error I have with the manually modified calibration file?


Regards
Andy

No, we would not do manual edits to improve the calibration results. We would rather reshoot the calibration video.
BTW, 5cm error is not that much and you still can get pretty nice tracking with such calibration. But with a poor calibration video, the error itself may be estimated incorrectly.

You can give a try to your manual calibration fitted to a T-pose frame. You cannot estimate the error in numbers, because this error is measured on an array of frames used for the calibration, as a difference between determined "average" value and values in specific frames. The best way to check whether calibration is good is to show the combined point cloud from all the cameras (menu "View > Show Depth From All Sensors") and check whether it looks more like an actor's body or disconnected pieces throughout the whole video.