One of the main challenges of Deep Learning derived from this is being able to deliver great performances with a lot less training data. As we will see later, recent advances like transfer learning or semi-supervised learning are already taking steps in this direction, but still it is not enough .