Have you ever stumbled upon an image (or meme) you really love from the internet that has poor quality and is blurry, and even Google couldn't help you to find a high-resolution version of it? Unless you are one of the few who have spent years learning math and coding, knowing exactly which fractional-order regularization term in your objective equation can be solved by which numerical method, we might as well give GANs a shot!
This chapter will help you to perform image super-resolution with SRGAN to generate high-resolution images from low-resolution ones and use a data prefetcher to speed up data loading and increase your GPU's efficiency during training. You will also learn how to implement your own convolution with several methods, including the direct approach, the FFT-based method, and the im2col method. Later on, we will get to see...
