mnist_autoencoder_cnn.py

from keras.layers import Input, Dense, Conv2D, MaxPooling2D, UpSampling2D
from keras.models import Model
from keras import backend as K
import matplotlib.pyplot as plt

def get_ae_cnn(nb_filter=32, kernel_size=3, max_pooling_size=2 ):
    input_img = Input(shape=(28, 28, 1))  
    x = Conv2D(nb_filter, (kernel_size, kernel_size), activation='relu', padding='same')(input_img)
    x = MaxPooling2D((max_pooling_size, max_pooling_size), padding='same')(x)
    x = Conv2D(nb_filter, (kernel_size, kernel_size), activation='relu', padding='same')(x)
    encoded = MaxPooling2D((max_pooling_size, max_pooling_size), padding='same')(x)
    encoder = Model(input_img,encoded)	
    x = Conv2D(nb_filter, (kernel_size, kernel_size), activation='relu', padding='same')(encoded)
    x = UpSampling2D((max_pooling_size, max_pooling_size))(x)
    x = Conv2D(nb_filter, (kernel_size, kernel_size), activation='relu', padding='same')(x)
    x = UpSampling2D((max_pooling_size, max_pooling_size))(x)
    decoded = Conv2D(1, (kernel_size, kernel_size), activation='sigmoid', padding='same')(x)
    autoencoder = Model(input_img, decoded)
    return autoencoder, encoder
    
def fit_ae_cnn(ae, x_train_noisy, x_train, x_test_noisy, x_test, epochs=5, batch_size=128):
  ae.compile(optimizer='adadelta', loss='binary_crossentropy')
  early_stopping = EarlyStopping(monitor='val_loss',min_delta=0,patience=5,verbose=1, mode='auto')
  filepath = "saved-model-{epoch:02d}-{val_loss:.2f}.hdf5"
  checkpoint = ModelCheckpoint(filepath, monitor='val_loss', verbose=1, save_weights_only=True, save_best_only=False, mode='max')
  history = ae.fit(x_train_noisy, x_train,
                          epochs=epochs,
                          batch_size=batch_size,
                          shuffle=True,
                          validation_data=(x_test_noisy, x_test), 
                          callbacks=[early_stopping, checkpoint])
  return history

def plot_training_loss(history):

  loss = history.history['loss']
  val_loss = history.history['val_loss']

  epochs = range(1, len(loss) + 1)

  plt.plot(epochs, loss, 'bo', label='Training loss')
  plt.plot(epochs, val_loss, 'r', label='Validation loss')
  plt.title('Training and validation loss')
  plt.xlabel('Epochs')
  plt.ylabel('Loss')
  plt.legend()

  plt.show()

autoencoder_32, encoder_32 = get_ae_cnn(nb_filter=32)
history_32 = fit_ae_cnn(autoencoder_32, x_train_noisy, x_train, x_test_noisy, x_test, epochs=20)
plot_training_loss(history)