Fully Unsupervised Probabilistic Noise2Void

This is an interesting new pre-print from Jug lab: https://prelights.biologists.com/highlights/fully-unsupervised-probabilistic-noise2void/. The labs has been working on using Deep Learning based de-noising already for some time and originally co-authored the CARE method. Importance of demonising for microscopy and Life Sciences cannot be overstated. It literally allows to take a rather poor signal and get a very sharp signal out of it.

What they do here is to propose a self-supervised training scheme for de-noising. Has anyone DL based de-noising in their images?

Their source code is here https://github.com/juglab/ppn2v

I hope you mean denoising and not demonising! Or have I stumbled upon the occult dark-web branch of AI?!

1 Like

Indeed :joy: although the paper does seem to involve some dark magic :thinking:

1 Like