2x model about pytoflow HOT 4 CLOSED

Hello.

I remembered it was trained to complete 2x super-resolution.
But anyway, of course the algorithm can be applied to any-x super-resolution.
For example, you can upsample to 2^{n}-x picture with TOFlow and then downsample it to m-x picture (m < 2^{n}).

You can get more information and have a deeper understanding after reading the paper.

Best,
Coldog

from pytoflow.

I am sorry, it does perform 4x super-resolution since it transforms [n, m] to [2n, 2m].

In this way, I want to confirm what do you mean the "2x super-resolution"?

from pytoflow.

Hi @Coldog2333

Thank you for the response! I think you have got my point. By saying 2x super-resolution, I mean the output image is 2x in size than the input image, i.e. transforming a n x m image to 2n x 2m.

Now, the code is actually performing 4x resolution, where the output is 4x in size than the input (as also indicated in the paper), i.e., n x m to 4n x 4m. I wonder whether it is possible to revise the code for 2x super-resolution. Thank you!

Best,
Yongcheng

from pytoflow.

Hello,

Yes, sure. The size is not the problem.
Checking Issue #11 may be helpful for you.

Best,
Coldog

from pytoflow.