ashawkey / torch-ngp

A pytorch CUDA extension implementation of instant-ngp (sdf and nerf), with a GUI.

Geek Repo:Geek Repo

Github PK Tool:Github PK Tool

Sometimes, start from white result

jaepoong opened this issue · comments

I trained hash nerf at Photoshape dataset using script
python main_nerf.py data/photoshapes/shape09096_rank02 --workspace workspace/trial_02 -O

But when i trained, in many cases, it start with whitened result like below result with low PSNR
image
image

After training progress, we can get stable result like show,
imageimage

I don't know why this occur. This phenomenon occur in many scene at dataset.
Does anybody know why this happen??

somtimes, there is never trained case