k_umr_mri/blurAE/notes.txt

# 131,269 params
# 64,197 params
16,581 params

20 * 8739 = 174'780 training data


best loss during train (mse): 0,0285853