This is the notebook in which i want to solve the task of removing Gaussian blur from a set of images, so we need a good features extractor and denoiser. There are many models that are able to remove the noise from images, but since we are dealing with Gaussian noise, we will focus on Convolution and Convolution Transposed (also known as Deconvolution) based Neural Networks. In particular, i will show some results computed with a simple autoencoder architecture and with a REDNET one, that is the main model of the notebook.
The notebook is completely runnable and trainable. I you just want to check the predictions or results, there are the weights available on the repository.