Hi my friend,
By choosing the AI model you can also choose whether to upscale 2x or 4x.
Input resolution percentage is a function to downscale the image or video before passing it to AI.
For example, 500x500 image, FSRCNN_X4, Input resolution 50%.
=> 500x500 => 50% (500x500 = 250x250) => FSRCNN_X4 (250x250 x 4=1000x1000)