Keras Deep Learning Cookbook

Initializing ImageDataGenerator

Use keras.preprocessing.image.ImageDataGenerator(featurewise_center=False, samplewise_center=False, featurewise_std_normalization=False, samplewise_std_normalization=False) to initialize ImageDataGenerator.

The parameters are explained as follows:

• featurewise_center: Boolean. Set input mean to 0 over the dataset, feature-wise.
• samplewise_center: Boolean. Set each sample mean to 0.
• featurewise_std_normalization: Boolean. Divide inputs by std of the dataset, feature-wise.
• samplewise_std_normalization: Boolean. Divide each input by its std.

Let's apply this to our mnist dataset:

from keras.preprocessing.image import ImageDataGeneratorfrom keras import backend as KK.set_image_dim_ordering('th')

X_train = X_train.reshape(X_train.shape[0], ...