search

NVlabs / nvdiffrec

Official code for the CVPR 2022 (oral) paper "Extracting Triangular 3D Models, Materials, and Lighting From Images".
Other
2.09k stars 222 forks source link

Use implicit model #41

Open derrick-xwp opened 2 years ago

derrick-xwp commented 2 years ago

Hi

the mesh seems be explicitly learned, how to modify the config for implicit learning?

Thanks

jmunkberg commented 2 years ago

I'm not sure I understand the question, can you provide more details?

LZleejean commented 2 years ago

I think he expect that the sdf value is predicted by a neural network.

derrick-xwp commented 2 years ago

I'm not sure I understand the question, can you provide more details?

Using MLP to predict SDF. Is the code only a partial implementation of the original paper?

jmunkberg commented 2 years ago

Yes, we didn't include the DTU experiments from our supplemental material with the MLP prediction of the SDF in the public code release.

Some additional details in this issue: https://github.com/NVlabs/nvdiffrec/issues/9

Tom94God commented 2 years ago

Could you open the code with the MLP prediction of the SDF? It seems that the whole framework have to be changed when using MLP to predict SDF.

jmunkberg commented 2 years ago

No, we are not planning to release that part of the code unfortunately.

Implementation details are included in Section 8.5 and 9.2 of our paper. https://nvlabs.github.io/nvdiffrec/assets/paper.pdf It is mostly the dmtet.py file that needs to be modified.

We use the same MLP as in Differentiable Volumetric Rendering: Learning Implicit 3D Representations without 3D Supervision, with code available here: https://github.com/autonomousvision/differentiable_volumetric_rendering