A CNN example with Keras and CIFAR-10

In Chapter 3, Deep Learning Fundamentals, we tried to classify the CIFAR-10 images with a fully-connected network, but we only managed 51% test accuracy. Let's see if we can do better with all the new things we've learned. This time we'll use CNN with data augmentation.

1. We'll start with the imports. We'll use all the layers we introduced in this chapter, as shown in the following example:

import keras
from keras.datasets import cifar10
from keras.layers import Conv2D, MaxPooling2D
from keras.layers import Dense, Dropout, Activation, Flatten, BatchNormalization
from keras.models import Sequential
from keras.preprocessing.image import ImageDataGenerator

2. We'll define the mini batch_size for convenience, as shown in the following code:

batch_size = 50

3. Next, we'll import the CIFAR-10 dataset and we'll normalize the data by dividing it to 255 (maximum pixel intensity), as shown in the following code:

(X_train, Y_train), (X_test, Y_test) = cifar10.load_data()

X_train...