Level Up Moments (for Checkpoint)

The entire process of JPG was new, but the most interesting aspect was how the DCT is applied to an entire 2D image to produce the desired matrix and preparations for quantization. This was perhaps the clearest demonstration of an application of what we learned in class (FFT) and how it applied to a real world situation. It was also novel to experience how a change in basis could be applied not just to traditional 1d signals like a noise input, but a 2D image,

Image compression using machine learning techniques is not something we have considered in the past. Autoencoders are a somewhat simple machine learning approach to compression, but learning about other approaches using GANs was very interesting.