Open Feynman1999 opened 3 days ago
Hi! I have not published a 'formal' analysis on both datasets, unfortunately. However, empirically, it gives better metrics. We have discussed a bit about it on issue 86. @Phhofm will publish an article soon comparing DF2k against distilled versions, so if you're interested in this subject he might be able to share some data with you.
Hello! have you compared the performance of real-esrgan( RRDB blocks ) trained on these two datasets?