Black image result about selftmo HOT 6 CLOSED

Hi, thanks for the attention. It's hard to tell what happen there from the limited information. However, do you check the intensities of the HDR images you use. If you use .exr images, from my experience, sometimes it will appear nagative values after do resizing.

from selftmo.

I'm using.exr; however, I've noticed that sometimes an identical image yields a black output, while other times it yields a standard one, so I think that it is not about intensities because some times the result is correct.

from selftmo.

Ok, I haven't encountered such problem. My suggestion is that to use a fixed seed for hyper parameters initialization.

from selftmo.

I can't find an exposed seed anywhere

from selftmo.

I mean to set a fixed seed, something like, torch.manual_seed(seed_value) in pytorch (you could check that in TF).

from selftmo.

Yep, thank you, that worked.

from selftmo.