Result degradation when training from one image

Question

Result degradation when training from one image

Opened this issue 3 years ago · 1 comments

850207131 commented 3 years ago

Hi author, I try to train your work from one image that you deleted from the original Nerf. I use the same code but render worse results than training from the random batch. I really want to know whether you've faced this situation and I can't explain why it happened...

Answer 1 · 2022-03-11T07:32:18.000Z

Sorry, we have not try one image. All experiments are done with random batch so we delete one image part.