Been working on a color recovery machine learning model. I’m surprised how well Pix2PixHD is working (obviously looking at test set data so caveats there). First image is input, faded print. Second is the output from PIx2PixHD. 3rd is ground truth (what is should look like)