Data Augmentation

Data Augmentation

from keras.preprocessing.image import ImageDataGenerator

datagen = ImageDataGenerator(
        featurewise_center=False,  # set input mean to 0 over the dataset
        samplewise_center=False,  # set each sample mean to 0
        featurewise_std_normalization=False,  # divide inputs by std of the dataset
        samplewise_std_normalization=False,  # divide each input by its std
        zca_whitening=False,  # apply ZCA whitening
        rotation_range=15,  # randomly rotate images in the range (degrees, 0 to 180)
        zoom_range = 0.1, # Randomly zoom image 
        width_shift_range=0.1,  # randomly shift images horizontally (fraction of total width)
        height_shift_range=0.1,  # randomly shift images vertically (fraction of total height)
        horizontal_flip=False,  # randomly flip images
        vertical_flip=False)  # randomly flip images

datagen.fit(x_train)

Make Model

X = tf.keras.layers.Input(shape=[28, 28, 1])
H = tf.keras.layers.Conv2D(32, kernel_size=5, padding='same', activation='swish')(X)
H = tf.keras.layers.MaxPool2D()(H)

H = tf.keras.layers.Conv2D(32, kernel_size=5, padding='same', activation='swish')(H)
H = tf.keras.layers.MaxPool2D()(H)

H = tf.keras.layers.Conv2D(32, kernel_size=3, padding='same', activation='swish')(H)
H = tf.keras.layers.MaxPool2D()(H)

H = tf.keras.layers.Conv2D(64, kernel_size=3, padding='same', activation='swish')(H)
H = tf.keras.layers.MaxPool2D(pool_size=(2, 2), strides=(2, 2))(H)
H = tf.keras.layers.Dropout(0.3)(H)

H = tf.keras.layers.Flatten()(H)

H = tf.keras.layers.Dense(128)(H)
H = tf.keras.layers.BatchNormalization()(H)
H = tf.keras.layers.Activation('swish')(H)

H = tf.keras.layers.Dense(84)(H)
H = tf.keras.layers.Activation('swish')(H)
H = tf.keras.layers.Dropout(0.5)(H)

Y = tf.keras.layers.Dense(10, activation='softmax')(H)
model = tf.keras.models.Model(X, Y)
model.summary()

train_test_split

from sklearn.model_selection import train_test_split

x_train, x_val, y_train, y_val = train_test_split(x_train, y_train, test_size=0.1)

Model fit

batch_size = 86

history = model.fit(datagen.flow(x_train, y_train, batch_size=batch_size),
                              epochs = 30, validation_data = (x_val,y_val),
                              verbose = 2, steps_per_epoch=x_train.shape[0] // batch_size)