I have an optimized SSM and want to generate and save some meshes using the Python API.
From reading this thread on surface reconstruction by warping the mean mesh, I see that the MeshWarper class is available to generate meshes based on a reference mesh and particles, but I am confused about what mesh I should use as the reference. One option would be to use the mesh I used as a reference for my alignment step. Otherwise, I guess really any of the subject meshes could be used. But what is the recomended approach? I have also read some mentions of a “mean mesh” but can’t find a ShapeWorks function to create this.
Got it, thanks for clearing that up. By the way, I plan to publish my model in such a way that others can create meshes from it. Due to the access agreement of the dataset I am using, I must not publish any of the individual meshes used to create the model. In order to get a publishable reference mesh for warping, I guess I can use the median to generate the mean mesh, which I can publish along with its particles and use for all subsequent warping.
As for the remaining information needed to replicate the model, am I correct that it can be reconstructed from just the eigenvectors, eigenvalues, and PCA scores? The scores themselves I may also not be able to publish since they could be used to create a representation of a real individual subject, but perhaps I could publish the standard deviations (with the means being zero in my understanding).
Yes, that’s correct. The mean mesh, eigenvectors and eigenvalues should be sufficient. The PCA scores can be used to re-create the original shapes, so I would exclude those.