Keras

Keras is the most high-level deep learning library available, and is often where people begin on their AI journey. While we will focus on applications with TensorFlow in this book, it is important to introduce Keras because of its ubiquity and ease of use. 

Written by Fran?ois Chollet at Google, Keras is a wrapper that can run on top of TensorFlow or other libraries such as Apache, MXNet, or Theano. Like the other libraries, it is available through PyPy by running pip install keras in your terminal or command line. Functionally, it's very similar to the way the scikit-learn works, and hence is a popular library for those who wish to get their hands dirty with deep learning as quickly as possible. 

Like PyTorch, Keras was designed to be pythonic; everything utilizes native Python functions. It was also designed to be modular and minimal so that code is easily portable and reusable. To run Keras, you must already have TensorFlow installed, as it will act as it's back-end operator.