Google Code-in

Data augmentation is a powerful technique that helps to prevent overfitting. It is a commonly used technique in tasks like data augmentation and is particularly found to be very useful when the volume of data is less. This tutorial provides a decent overview of data augmentation using Keras's ImageDataGenerator class.

The objective of this tutorial is to create a tutorial that shows how to build flexible data augmentation pipelines using TensorFlow's image module which is popularly referred to as tf.image. Optionally, it should also show how to use such a pipeline on the fly and train a model (built using tf.keras).