search

VITA-Group / SinNeRF

[ECCV 2022] "SinNeRF: Training Neural Radiance Fields on Complex Scenes from a Single Image", Dejia Xu, Yifan Jiang, Peihao Wang, Zhiwen Fan, Humphrey Shi, Zhangyang Wang
MIT License
328 stars 26 forks source link

I want to ask how to prepare your own data set? #5

Open Hzurang opened 2 years ago

Hzurang commented 2 years ago

I want to ask how to prepare your own data set? Thank you.

Ir1d commented 2 years ago

What do you mean by your own dataset?

LitterWindwind commented 2 years ago

Train on custom dataset?

SEJUNHONG commented 2 years ago

I want to know how to make depth files(.npy files)??

Ir1d commented 2 years ago

.npy files contain depth values @SEJUNHONG

Wanggcong commented 2 years ago

Hi, I have a question about how to get depth maps (or .npy files in this repo) given RGB images?

HannahHaensen commented 1 year ago

@LitterWindwind are the depth files obtained using monocular depth estimation or e.g. for an own dataset using Blender for example to generate it, directly RGB-D values can be used?

Blender can export this for sure, my question is just is it possible or for a better comparison another techqique should be used?

so the main question is how is the depth generated for the exisiting datsaets?

Ir1d commented 1 year ago

I will release updates to train SinNeRF with monocular depth soon. Currently, for NeRF synthetic and LLFF, we use a pre-trained NeRF to obtain the depth. For DTU, we use MVSNet_pytorch to obtain the depth.