Does mesh connectivity matter for articulation prediction?

[tcm03]

Hi,
Thanks for sharing your impressive work on articulated object estimation with a lightweight feed-forward approach!
As a matter of fact, an object might have submeshes for its parts separated from each other, or just one entire mesh. My question is, does the connectivity of the object's mesh matter for the model to predict?
(my thought is that since the input point cloud is sampled from the mesh anyway, whether it is one whole mesh or collection of meshes for parts, doesn't matter, but it would be great to hear from the authors)
Thanks

[RuiningLi]

Hi,

Thanks for your interest in our work! The mesh connectivity does NOT affect the network prediction as you said! However, we have a post-processing procedure which converts the network prediction to the final articulated asset which gets stored eventually. We do leverage the connected components of the original mesh as a clue during post-processing! This part is implemented here if you want to take a closer look!