Dear OpenSim-Team, dear Forum-Users,
First of all, I wanted to thank you for the nice work you are doing keeping OpenSim an open source product and putting so much effort in the development of this tool.
A colleague and I are currently working on Batch processing through our data with Matlab. Everything is working fine we only have some problems getting the RMS errors of the scaling for each Marker. We know that the RMS for the average and the max RMS are being printed within the command window of OpenSim. But we were wondering, if there is a possibility to store the RMS for each marker within a file, so we can build a pipeline to compare our scaling results without manually clicking in the OpenSim GUI.
The second question is regarding the picture I attached to this email. So far scaling with a restricted model in knee DoFs is not a problem. We tried to use the same settings for a 6 DoFs knee model, but what happens is that the bones are getting disorganized as you can see in the attached file. I know that joint constraints are a strength of OpenSim, nevertheless it would be nice to use also a 6DoFs model for our purposes.
Thank you for your help.
//Markus
RMS errors of scaling for each Marker/Scaling 6 DoFs knee model
-
Markus Kurz
- Posts: 9
- Joined: Mon Jun 01, 2020 10:37 am
RMS errors of scaling for each Marker/Scaling 6 DoFs knee model
- Attachments
-
- complex_model.JPG (26.94 KiB) Viewed 418 times
Tags:
-
Colin Smith
- Posts: 53
- Joined: Fri Feb 24, 2012 11:50 am
Re: RMS errors of scaling for each Marker/Scaling 6 DoFs knee model
I'm not sure about a nice solution for the first question through the API.
For the second question, the scale tool is essentially performing inverse kinematics for your input .trc marker data. So an optimization is being solved to calculate the joint angles that pose the model to minimize the errors between the virtual markers fixed to the model and the experimental marker data at one time step.
If you make the knee a 6 DOF joint, errors in marker placement and segment scaling etc will mean you likely will never get the knee to line up perfectly, but what you are showing is far too big to be this sort of error. My guess is that you have the right and left naming of one of the knee markers in the experimental data flipped (or could be the markers on the model are mislabeled). So essentially the right knee marker in the model is trying to be matched to the left knee marker in the experimental data. This might get hidden if you have a 1 DOF knee, since only the knee flexion angle is solved for, and the weightings of the other correctly labeled markers on the leg might prevent the hip angles from looking very strange.
For the second question, the scale tool is essentially performing inverse kinematics for your input .trc marker data. So an optimization is being solved to calculate the joint angles that pose the model to minimize the errors between the virtual markers fixed to the model and the experimental marker data at one time step.
If you make the knee a 6 DOF joint, errors in marker placement and segment scaling etc will mean you likely will never get the knee to line up perfectly, but what you are showing is far too big to be this sort of error. My guess is that you have the right and left naming of one of the knee markers in the experimental data flipped (or could be the markers on the model are mislabeled). So essentially the right knee marker in the model is trying to be matched to the left knee marker in the experimental data. This might get hidden if you have a 1 DOF knee, since only the knee flexion angle is solved for, and the weightings of the other correctly labeled markers on the leg might prevent the hip angles from looking very strange.
-
Markus Kurz
- Posts: 9
- Joined: Mon Jun 01, 2020 10:37 am
Re: RMS errors of scaling for each Marker/Scaling 6 DoFs knee model
Hej Colin,
Thanks for you reply.
I have double checked the marker naming in the experimental as well as in the virtual marker data set. They both fit perfectly. I solved it by deleting the translational DoFs because rotational (ext-flex, add-abd, rot) is most important for us. But its still weird getting those kind of results when not constraining the translation. The weighting of the marker I did according to the "guidelines" of OpenSim and our overall RMS is also quit good.
Best regards,
/Markus
Thanks for you reply.
I have double checked the marker naming in the experimental as well as in the virtual marker data set. They both fit perfectly. I solved it by deleting the translational DoFs because rotational (ext-flex, add-abd, rot) is most important for us. But its still weird getting those kind of results when not constraining the translation. The weighting of the marker I did according to the "guidelines" of OpenSim and our overall RMS is also quit good.
Best regards,
/Markus