csxmli2016 / redegnet Goto Github PK

Thank you for providing such excellent work. I am currently trying to reproduce your work. I wonder if you could provide me with a real-world facial dataset that you found through internet search?

Thanks for your excellent work. But when I use this to enhance some of my old photos, I observe some disappointment artifacts. Do you have any plans or ideas to improve it? (first is the origin image, second is the enhanced one, this is not the only one fail case, still more)

Your work is excellent! When will you share your work?

Hello.
Thanks for your great work and the code sharing.
I have several questions about the paper.

1. May I know the sr results that comparing with realesrgan bsrgan are in which upscale?
2. Can I simply replace the RRDB to other sr network, e.g. edsr?
3. I have tried the test_degradation, and it results in an image that with artifact and without a pixel downsampling, is that right?
4. Also, I notice that the real-world LQ/HQ in Train data contains around 10 images, is it need more if I want to retrain the network?
   Thanks a lot.