There’s this slightly forgotten repo on bitbucket with an image synthesis project for use in machine learning. Basically you add a script to a camera object and it renders to additional ‘passes’ to consecutive camera display ‘channels’, things like object ID, category ID (assigned by layers), depth, screen normal etc. The whole thing is based on custom shaders (one for optical flow and one for all the other ‘channels’) which were written before SRP and obviously don’t work in HDRP.
My question is if it’s doable in HDRP at all and if so can it be done with the same approach, or does it require a different set-up.
You can find the shaders under this link:
https://bitbucket.org/Unity-Technologies/ml-imagesynthesis/src/master/Assets/ImageSynthesis/Shaders/
Were you able to find a solution?
@andrzej_1 Could you find any solution for HDRP?