search

iMoonLab / MeshNet

MeshNet: Mesh Neural Network for 3D Shape Representation (AAAI 2019)
MIT License
341 stars 59 forks source link

FRC #22

Closed yhl48 closed 2 years ago

yhl48 commented 2 years ago

Hi, could you clarify the purpose of the structural descriptor? It looks like FRC would also model the volume of a 3D object given that the inputs to FRC are the corners, but at the same time you mentioned that it would capture the inner structure of an object, which would be 2D. Is it wrong to assume that that part of the model would capture the volumetric information?

thss15fyt commented 2 years ago

Hi, thanks for your interest.

The structural descriptor is designed to capture features of local geometry on mesh without considerring the absolute spatial position of faces. The FRC takes the relative coordinates corners - center as input, and models the structural information of each polygon face in 3D space.