Summary
Keras makes it remarkably easy to build and train neural networks to perform a wide range of deep-learning tasks. It doesn't absolve you from understanding various neural network architectures, knowing what types of layers to include in a network or the sizes of those layers, or, for example, understanding the role of activation functions and when to apply activation functions of different types, but once you've determined what to build, a few lines of code generally gets the job done. Moreover, Keras allows you to experiment with different network architectures and implementations and rather quickly compare the results.
Learn more
If you would like to learn more about Keras, there are some great online resources available, including the Keras blog. You may also want to follow François Chollet on Twitter. François is the author of Keras as well as the author of the book Deep Learning with Python, which is the ultimate self-learning guide for Keras. In addition to introducing Keras and providing key insights to help you use it effectively, the book provides a working introduction to deep learning that is independent of the platforms and libraries that you use.