search

zju3dv / object_nerf

Code for "Learning Object-Compositional Neural Radiance Field for Editable Scene Rendering", ICCV 2021
https://zju3dv.github.io/object_nerf/
MIT License
313 stars 23 forks source link

Scene-guided occlusion identification #10

Closed yixchen closed 2 years ago

yixchen commented 2 years ago

Hi,

Thanks for the paper and the code contribution. Nice work!

One question: could you please point out the code reference for the "scene-guided occlusion identification (scene guidance + 3D guard mask)" and how they affect the loss backpropagation? Thanks again.

ybbbbt commented 2 years ago

Hi, yixchen,

The code snippet related to 3D guard mask is:

https://github.com/zju3dv/object_nerf/blob/e8b1a7a5ab0596babdf4400a9e8908f1bfdcf990/models/rendering.py#L192-L202

As for the scene guidence, since we always use weights_coarse for both scene and object sampling, it directly inherits sampling distributions from the scene branch:

https://github.com/zju3dv/object_nerf/blob/e8b1a7a5ab0596babdf4400a9e8908f1bfdcf990/models/rendering.py#L305-L310

yixchen commented 2 years ago

Hi, thanks for the prompt reply. I will look into them.