Page 1 of 1

Scale and IK Problem - solved :-)

Posted: Mon Nov 22, 2021 8:05 am
by lisa-jonas
Hi at all,

I am working on an university-project to evolve a workflow for transfering mocap-data (optitrack) to OpenSim and afterwards to detect malpositions with AI (trained with biomechanics data). At the moment I am working with the raabe full body model and self recorded moCap-Data (with an Optitrack-MoCap-System and a Markerset with 49 Markers). I had to modify the markers of the "raabe full body model" a bit to fit to my recorded data (number, positions and names).

Now to my actual problem:
When I try to scale the model on my moCap-data or to use the IK-Tool, the model does not adopt the correct position (a standard T-pose).
Especially hip, pelvis and torso orientations are wrong, when I run IK, as you can see in the attached picture "RaabeFullBodyModelNotAdeptingT-Pose_red.png".

Own reflections on the possible cause of the problem:
My self recorded data imported in OpenSim "shows the subject looking to the back". Is that the problem? Is the model not able to adopt a position looking to the back?
(Using the coordinates-settings I realized that the pelivs-rotation can only be set +-90 degree. For for my collected data I think I would need a pelivs-rotation of +-180 degree.)

Thank you very much for any help.

Regards
Lisa Jonas

PS:
I also tried to attach the modified model, but I can't because its size is to big.
In the hope that it could also be helpful, I have attached a file with the recorded data of the T-Pose (a trc file change to txt, because this forum doesen't accept the upload of trc) and the IK-Results-file.

Re: Scale and IK Problem

Posted: Mon Nov 29, 2021 7:24 am
by lisa-jonas
Rotating the markermeasurments 180 degree (before importing into openSim) solved the problem and the model adopts the T-Position correctly.
(The rotation makes the subject (of the imported data) look to the front, instead of to the back)

Re: Scale and IK Problem - solved :-)

Posted: Mon Nov 29, 2021 9:47 am
by aymanh
Thanks for reporting, Lisa.

If you have lots of data and the model can actually move in arbitrary direction, it may be easier to change the bounds of the pelvis rotation coordinate and have the Scale/IK rotate the model for you instead.

Hope this helps,
-Ayman