In this repo you can find the Python Notebook and a brief report written in Italian.
The assignment consists on the prediction of grayscale images of letters P - Z of handwritten letter dataset with a visual investigation of the reconstruction abilities of an auto-encoder architecture.
The Dataset consists of 14000 training labelled images and 8800 test images.
The Analysis has been conducted in 3 steps: first I have explored the dataset and made the preprocessing operations needed; second the Neural Network has been compiled and evaluated; last, the autoencoder's reconstruction abilities have been investigated and a Neural Network with encoded input has been designed, evaluated and compared with the baseline one.
