Training image sizes

[ganav]

Hello i am using my own code and own dataset for super-resolution.

I have a question that in "Perceptual Losses for Real-Time Style Transfer and Super-Resolution: Supplementary Material" in the Table2, architectures take smaller image for input and generate bigger image for output. Means that smaller is low-resolution and bigger is high-resolution.

If input size of architecture is 72x72 then how could low and high-resolution images in training have same size (72x72) when output is 288x288?

This is confusing me how to make my training sets. Thank you