can we use nerf for llff dataset?

[qingxu-thu]

How can we use the llff dataset as origin nerf for some results, I wonder if some interfaces for the llff input, since I'm not sure whether there can be a conving transform for llff poses to the transforms_train?

[Gword]

You can modify dataset.py to support llff dataset. For reading llff data, you can refer to: https://github.com/Jittor/jrender/blob/main/nerf_helper/load_llff.py

[qingxu-thu]

Ok, It seems the implementation is corresponding to the original LLFF. It can try to use its c2w matrix. But I see the original instant ngp collmap2nerf.py and for c2w, it seems to conduct a coord transform for the origin. I wonder if it has been done in this implementation.

[Gword]

We support using llff dataset in the latest version. You can pull the latest version and have a try.