my datasets were collected from real optical device. the Magnification between input and target are 3.2.
so if I want to design deep learning models to learn the mapping from input to the corresponding target, how should I design the upsample layers?