I recommend to use the PyTorch code for training and testing. The model parameters of MatConvnet and PyTorch are same. - maintrainsrmd.py - maintestsrmd.py
Features
Computer Vision Research - Single convolutional network for multiple super-resolution degradations.