I wonder if you happen to know who is currently the number one in significantly enhancing photo resolution by “drawing” missing details through Machine Learning? I know about ESRGAN, GFPGAN, but they work quite poorly. It would be interesting to find out why there isn’t remarkable progress in this area? After all, there are many high-resolution faces, or a lot can be done, and it seems quite logical to train the system on a before-and-after basis. The result of ESRGAN from the first blurry picture in the comments is really far from what is desired.
Hi, I'm Rauf Aliev. 