convolutional neural networks (CNNs) based solutions have achieved
state-of-the-art performances for many computer vision tasks, including
classification and super-resolution of images. Usually the success of these
methods comes with a cost of millions of parameters due to stacking dee