About the benefits of NeuMesh over the extracted geometry from neural fields

[07hyx06]

Hi, thanks for the great work and code.

I have a question about this work. To support texture editing or geometry editing, a naive solution is to train a NeuS on the captured multi-view images, extract the mesh from the NeuS, and perform UV unwrapping. This way, we can import the assets to a 3D CG software, e.g. Blender, and perform all kinds of geometry and texture editing shown in the NeuMesh paper. Does NeuMesh achieve better rendering quality than the naive solution, or has other benefits?

[chobao]

Hi, 07hyx06, Thanks for your interest in our work!

I think extracting textured mesh from NeuS and rendering it by the 3D CG software will induce less photo-realistic results. The main reason is that the mesh texture is poor and loses the view-dependent phenomenon. Though some NeRF products (e.g. Luma AI) have been working on extracting photo-realistic mesh from NeRF, vanilla NeuS can not satisfy this demand.

A further discussion is also included in "Section C. Using neural implicit representation instead of traditional textured mesh" with Figure K in our supplementary material.

Hope it will help you!