2x resolution instead of 4x

[ShubhamCpp]

Hi,

First of all, thanks.
I'm getting good results for 4x upscale factors.
However, I wanted to know what changes would I have to the model to prepare a 2x upscale model.

I know that one can always downscale. However, I'm hoping to get better performance (in terms of accuracy and speed) if I change the model to resolve for a 2x upscale factor.

Thanks,
Shubham

[brade31919]

Hi,
In this case, you need to modify the network structure and do the re-training. It's a little bit stupid I know, but that's the structure of SRGAN (If I modify the structure and simultaneously train the net on 2x and 4x tasks, the result might be even better but it's not the original SRGAN anymore.)

Best,
brade31919