This is a brief project created for the course of Statistical Foundation of Machine Learning. In this project I created a CNN able to correctly classify photos of cats and dogs. I tried several architectures and then choose the best one and run a CV on it to compute the risk estimate. In the end the accuracy of my CNN has been of 87.86%(+/- 1.43%).