Keras is an API that makes using and building deep learning models easier and faster. So it's a deep learning toolbox that's all about:
- Ease of use
- Reducing complexity
- Reducing cognitive load
And by making deep learning easier to use what happens is that you are making it accessible to more people. So the key design concept behind Keras is that to put deep learning into the hands of everyone.
So Keras is more like an API that has several different implementations. There's the Theano implementation which was originally released with Keras and you also have the TensorFlow implementations, and more will be coming in the future.
What TensorFlow has done is bringing the Keras API into the TensorFlow project and it's a great to combine the advantages of both TensorFlow and Keras and making deep learning accessible to everyone.