In Single Image Super Resolution (SISR), we have two paired sets of low resolution and corresponding high resolution images. Suppose we have a set of high quality images of the size 512 x 512 x 1 and its corresponding set of low quality images of the size 64 x 64 x 1. The purpose of the SISR is to feed the low-resolution images as the inputs to the model while the high resolution images are the labels. Then, after training has finished, the neural network model learns to take a low resolution image of the size 64 x 64 x 1 and produce its corresponding high resolution image of the size 512 x 512 x 1. Now, my question is that:
is it possible to train this model in such a way that it takes the low resolution images 64 x 64 x 1 as inputs, and produce 265 x 256 x 1 and 512 x 512 x 1 images simultaneously, while the only labels is still the high quality images of the size 512 x 512 x 1?
I tried to explain my question as much as clear as I could.
