I believe the reconstructions look like this because the particles are out of correspondence across shapes. If you look at the PCA mode animations, there are likely particles swapping positions.
I would recommend enabling “normals” in optimizations, and perhaps “geodesics” to try to improve the correspondence.
Thank you for your reply. The link to the similar shapes you provided refers to data that I originally shared, so I am already familiar with those examples.
Compared to the dataset on Zenodo, I have since added more shapes, which introduce greater shape variation. Prior to reconstruction, I manually scaled the shapes to achieve equal circumference in order to reduce variation along that dimension.
At the link below, you can find the files I am currently using, along with the reconstructions I have obtained so far:
I was wondering whether you might be able to offer further guidance on how to optimize the reconstructed shapes. I believe the results could still be improved, as some of the reconstructions exhibit spikes or small holes.
Hereby a new link and yes I have tried the same parameters as you can see in the file. I just hope there is a way to improve them. Thanks in advance for taking time to have a look at it!
