Hi
first thanks for sharing that code, it is a very cool project.
I have a use case when I am optimizing the pose and shape parameters w.r.t. some animation loss function, but the model makes very unnatural appearance. Do you have a function or data which gives a statistical prior on the reasonable pose/shape parameters?
Hi @nikjetchev, the STAR kinematic tree is based on the SMPL kinematic tree. So you can re-use pose prior for SMPL (V-poser, GMM pose prior) also for STAR.
Hi first thanks for sharing that code, it is a very cool project.
I have a use case when I am optimizing the pose and shape parameters w.r.t. some animation loss function, but the model makes very unnatural appearance. Do you have a function or data which gives a statistical prior on the reasonable pose/shape parameters?
(e.g. as SMPL has, e.g. https://github.com/vchoutas/smplify-x/blob/master/smplifyx/prior.py )
Or at the very least, what is the numerical interval of the pose and shape parameters? Are the shape parameters actually angles in radians?
Or would it be a good idea to convert the old SMPL data into STAR format and calculate the prior myself?
thanks, any help will be appreciated