camera noise

[gpuartifact]

Hello,
Would you recommend to either:
a) add the expected camera noise to the simulation data for training to implicitly train de-noise
or
b) should the camera image be de-noised first as good as it can be before passing it to RAFT-stereo?

I am trying to train a for a specific camera sensor.

[lahavlipson]

This sounds like an empirical question. I suggest try different options and pick the model with the best validation performance.