Preprocessing Layers

Keras Preprocessing Layers are a set of Keras layers aimed at making preprocessing data fit more naturally into model development workflows. They can handle a wide range of input, including structured data, images, and text and can be combined directly with Keras models and exported as part of a Keras SavedModel.

Note

Keras input processing pipelines can also be used as independent preprocessing code in non-Keras workflows

With Keras preprocessing layers, you can build and export models that are truly end-to-end:

• models that accept raw images or raw structured data as input

• models that handle feature normalization or feature value indexing on their own.

To learn more about Keras Preprocessing layers, review the videos in the following sections.

Notebooks & Documentation

• Keras Preprocessing Guide

• Structured data classification with preprocessing layers

• Sentiment analysis with preprocessing layers

• TensorFlow tutorial: Load and preprocess text

• TensorFlow tutorial: Load and preprocess images

Quick introduction

Check out this quick introduction to machine learning data processing as well as its challenges.

Learn how to easily prepare your data using the new Keras Preprocessing Layers API – in particular, how to do asynchronous preprocessing as part of your data pipeline, and how to export an end-to-end model that embeds its own preprocessing logic:

End-to-End example

Google Software Engineer Matthew Watson highlights Keras Preprocessing Layers’ ability to streamline model development workflows. Follow along as he builds an end-to-end model showing what you can do with these layers.

Resources

• An Introduction to Keras Preprocessing Layers