search

huangjh-pub / di-fusion

[CVPR'21] [Jittor & Pytorch] DI-Fusion: Online Implicit 3D Reconstruction with Deep Priors
122 stars 13 forks source link

Changing resolution variable #6

Closed eriksandstroem closed 2 years ago

eriksandstroem commented 2 years ago

Thanks for sharing your code to this project.

It appears that you require that the variable resolution > 1 i.e. 2, 3 etc. Is there a way to make it work with resolution = 1? When resolution = 1, I would assume that the grid is sampled at the voxel size so if the grid consists of X voxels, then you sample X grid points. Please correct me if I am wrong.

heiwang1997 commented 2 years ago

Hi @tfy14esa

Yes, I guess at the high level this is correct. If you want to generate more fine-grained meshes, you can basically increase the resolution within each voxel.

As we are using large PLIVoxs, if you are using resolution = 1 then the extracted geometry would be way too coarse. Of course you can decrease the voxel size to compensate for this, but the power of implicit fields couldn't be exploited in that way.

Best, Jiahui

huangjh-pub commented 2 years ago

Closing due to inactivity. Feel free to open it again if the problem persists.