make encoder/decoder of different size and check them on single cells

[dalessioluca]

make encoder/decoder of size 28x28, 56x56, etc.
the architecture can be copied from https://github.com/AntixK/PyTorch-VAE/blob/master/models/vanilla_vae.py

[dalessioluca]

PROBLEM OF ANTIALIASING
READ MASK CNN TO SEE HOW THEY DO THE REGION PROPOSAL
maybe the problem is to have two transaformation in sequence (first downscaling and then cropping)
maybe the encoder should take the raw image not the outcome of the unet. Outcome of unet is good for mask probably.
test what happends in big code if encoder takes the raw image