Hands-On Convolutional Neural Networks with TensorFlow

Introduction to the tf.data API

Before continuing, we will take a look at TensorFlow's way of handling data input to any kind of model we might train. The TensorFlow tf.data API provides us with all the tools we might need to easily build complex input pipelines. One pipeline you might commonly build would involve loading raw training data, performing some preprocessing on it, shuffling, and then putting it into a batch ready for training. The tf.data API allows us to do all these steps in an easy way using simple and reusable pieces of code.

The tf.data API has two main components to it that you need to understand. The First is the tf.data.Dataset; this is what represents your raw data. More specifically, it holds a sequence of elements, ...