Scaling results - Evaluation

[Pia Stefanek]

Hi,
I performed Scaling with the MoBL_ARMS model using marker data from experiments. I read the page https://simtk-confluence.stanford.edu/d ... th+Scaling which explains results evaluation. It is written that the maximum marker error should be lower than 2cm and the RMS should be lower than 1cm. In my experiments there is the static pose missing. Therefore, I have to take a single frame from the motion file for scaling. In my scaling results I have maximum errors lower than 4cm and RMS lower than 3cm. I already adjusted the marker weights. Do you think there is anything more to do to improve my results or is it normal that I get larger errors since my static pose is missing?
Furthermore, on the page https://simtk-confluence.stanford.edu/d ... th+Scaling is written that one should adjust the virtual markers of the scaled model to match the experimental markers. Why is this the case?

Thank you very much.
Pia

[jimmy d]

I assume you mean this line from documentation;

After examining the "Messages" window and performing a visual comparison, adjust the virtual markers and marker weightings to improve your results. Again, avoid adjusting the positions of the landmark and FJC virtual markers to match the experimental markers.

What that is telling you is that if you feel that the scaled and registered model doesn't match your experimental subject, you will need to adjust the model markers and rescale/re-register. This makes sense if have a large mismatch between a model marker and an experimental marke-- you would see a poor scaling result or/and poor registration.

If you haven't already, you should check out the webinar on scaling we did a while back.

[Pia Stefanek]

Ok, thank you very much for your answer James.
I already saw the webinar and it helped me to understand more about scaling.
But again when do I need to achieve the recommended values:
maximum marker error lower than 2cm and RMS lower than 1cm? Directly after scaling the generic model or after marker adjustment of the scale tool?

I try to do it this way:
1.I put the markers on the locations where I think they are correctly positioned.
2.I perform scaling without marker adjustment and save the model.
3.I load the scaled model and put it into the scaling pose and also view the experimental marker positions. If there are high differences I adjust the markers.
4.I scale the model with marker adjustment. But here I never achieve values in the recommended range.
When I scale the model with adjusted markers again then I mostly get values in the recommended range.

Thanks for your help
Pia

[Pia Stefanek]

My problem is that I have experimental marker data from 8 persons performing a movement with the arm ( I use the MOBL_Arms model).
I do not have static marker data from the persons. The persons perform a crank turn movement. I have the data for the crank angle per time. So I selected a specific crank angle value (for example 100 degrees) and used this time frame as input for "Marker data at static pose" in the scale tool. Unfortunately, my resulting joint angles for the static pose derived from the scale tool vary a lot. I do not know how I can improve my results.
I thought of locking some coordinates but I do not know which because I do not have any information about the real joint angles during movement.

I am really thankful if anybody could help.
Thanks
Pia