Hi,
I wonder if anyone has come across deep learning work applied to photo auto focus or/and deblurring. Thanks! I found this paper, but wonder if there is anything more state of the art.
Off the top of my head: this is a regression problem, but it has a lot of overlap with depth inference. If your goal is to take an unblurred image and apply blur to it based on depth, then you’ll first need a depth mask. Given a depth mask, you don’t need deep learning to do the remaining step - existing postprocessing filters will do quite well, eg the ones in blender 3d. It’s just the depth mask that’s difficult. So I’d look into what the current state of the art is for converting from RGB to RGBD (rgb+depth).