Hello everyone
I am the author of this article<p>Image super-resolution is one of the most popular generative algorithm<p>A neural network takes a low resolution image and has to imagine & generate all the finer details<p>A lot of rapid progress has been made in this field coming from early stage ML models to recent TECOGAN<p>In this article I cover the task of super-resolution, the taxonomy of the algorithm developments, loss functions, performance metrics and popular datasets<p>In addition I have provided detailed code of ESPCN architecture to train a model from scratch and corresponding Paperspace gradient notebook<p>Also attached a pre-trained ESPCN model to play out with<p>Would be happy to discuss any further questions related to this topic