3D Similarity Clustering

Hi, maybe 3D-slicer with some extensions is the right tool, but I don’t know. I’m searching for input, hints, guidance, or other solutions to the following problem.

I have several thousand (possibly a couple hundred thousand) STL files of manufacturing parts. So, the approach should be scalable.

I want to cluster these parts by geometric similarity. Since the STL files have different orientations and scalings, an invariant approach would most likely make sense.

The clustering should be controllable by making clusters with at least 30 parts or having some similarity measuring ranging from 0% (all parts form one cluster) to 100% (every part is its cluster).

Is 3D-Slicer the right tool for such a task?

Not out of the box, no. It may be a platform where you could visualize results though.

You may find that an approach like this is helpful: [2205.15456] Registering Image Volumes using 3D SIFT and Discrete SP-Symmetry

I wouldn’t mind using 3D-Slicer as a toolbox, or code some scripts using it.

Thanks for the paper link. I’m unsure it fits what I want to do. Sounds pretty complicated.