about SIMPLE IMAGE SIGNAL PROCESSING USING GLOBAL CONTEXT GUIDANCE paper

[alexliyang]

this paper is very interesting， I hope to know when encoding the global guidence vector, which size to down the full image to input the Encode net ?