simtk.org

Hello,

I would like to process some all marker data for which I do not have any height and mass information.
You specify that approximate values are okay. How approximate is this? What if I set an average size of 170 cm and 70 kg?

Would it be a bad idea to decrease the weight of residuals to zero in the advanced options, so that only marker RMSE is taken into account?

While I'm at it, here is another related question.

Given that you optimize scaling and inverse kinematics at the same time, do you output a slightly different scaled model for each trial of the same subject?

In other words, if I set the approximate height and weight of all subjects to the same values, and use a consistent model and markerset across subjects, can I consider all my trc files as different trials of the same subject that I can click and drop? Or will it lead to one single scaled model, and therefore to flawed results?

-----
EDIT:
Okay I just answered myself by downloading data processed by another user: there is only one scaled model per subject, so it would not work.
Would you have a piece of advice for uploading lots of data from different subjects without it being too much of a hassle?

Hi David,

The height and mass are important parameters for the model scaling pipeline. The final height of the model is regularized based on the height that you provide, and the dynamics fitting step will use the mass you provide as an initial guess (but it can change to match the GRFs you provide).

AddBiomechanics will still try to fit the marker data you provide, but setting the wrong height could negatively affect your results, so try to record that information whenever you can.

Would you have a piece of advice for uploading lots of data from different subjects without it being too much of a hassle?

We are working on a command line interface for uploading large datasets to make this process simpler. Stay tuned for development updates.

-Nick

Thank you Nick for your answer.

I will get an approximate size from the marker data, and I guess the mass won't bear too much importance of I'm only interested in kinematics.

I'm looking forward for your command line interface!

Well I initially wanted to use AMASS marker data to combine accurate skeletons (with AddBiomechanics) with accurate body meshes (with MoSh++). It seems like we just got scooped
https://skel.is.tue.mpg.de/

That's life, we'll definitely use AddBiomechanics in any case!