search

autonomousvision / unisurf

[ICCV'21] UNISURF: Unifying Neural Implicit Surfaces and Radiance Fields for Multi-View Reconstruction
MIT License
417 stars 35 forks source link

Which method is applied to transform the density derived from NeRF into mesh? #23

Open YouyuChen0207 opened 11 months ago

YouyuChen0207 commented 11 months ago

Hello, thanks to your great work!

I would like to learn some detail in your experiment where you set NeRF as a baseline. You converted NeRF density into a mesh with a threshold. Did you apply Marching Cube or some other methods to achieve this? If you applied other methods, which one is your choice?

Thanks!

YouyuChen0207 commented 11 months ago

I found you applied MISE for mesh extraction of your work. Did you use the same method to performe experiment on NeRF with a threshold extracted occupancy? I just want to learn the details. Thanks!

robofar commented 4 months ago

Hello @YouyuChen0207 . I am also confused about this part? I don't understand how they extracted mesh from NeRF density output. Also they are mentioning that they used some density threshold when extracting a mesh, but I don't understand what is it used for. Can you explain me if you figured it out?

YouyuChen0207 commented 4 months ago

@robofar Hello, if you want to know how to extract mesh from NeRF density (may not be the same method in this paper), you could look up for Marching Cube, which is an available paper describing how to derive a mesh from volume density grid. Hope this is helpful for you.