About the view space!

[dwangxxx]

Hello, Thanks for your source code! I've read your paper recently, and I have one question about this paper.

1. You mentioned that you model the spatial query in the input view's camera space. This means that the poses is camera2world?
2. I want to know that how can I apply this method on my own data? Would you like to give me some suggestions? Thanks very much! ! ! Looking forward your apply!

[lagrange10]

Hi! friend from XJTU!

Have you solve this problem? I'm trying to use pixelNeRF in my project and I only have the data of LLFF form（which contains the pose file called pose_bounds.npy just like nerf-pytorch），but I don't find the preprocessing code that read LLFF data to pixelNeRF. Anyway, the data of DTU form, I have no idea about generate it by my own.

There is some possible solution that I figure out.

• write a scirpt to generate DTU's pose( cameras.npz ) (some matrix transform from colmap pose to the form of DTU)
• copy and modify the code that can read LLFF data from other NeRF project(e.g. nerf-pytorch, GeoNeRF)

But I have not try them for the busy daily life. I would appreciate it If you have solve it and tell me.

[andreamigliorati]

+1

[Shichao123123]

+1