​

// Encoder 

​

encoder = models.Sequential()

​

encoder.add(layers.Conv2D(32, 3, strides=1, padding='same', activation='relu', input_shape=x_train.shape[1:]))

encoder.add(layers.MaxPooling2D(2, strides=2))

​

encoder.add(layers.Conv2D(64, 3, strides=1, padding='same', activation='relu'))

encoder.add(layers.MaxPooling2D(2, strides=2))

​

encoder.add(layers.Conv2D(128, 3, strides=1, padding='same', activation='relu'))

encoder.add(layers.MaxPooling2D(2, strides=2))

​

encoder.summary()

​

// Decoder

​

decoder = models.Sequential()

​

decoder.add(layers.Conv2D(128, 3, strides=1, padding='same', activation='relu', input_shape=encoder.output.shape[1:]))

decoder.add(layers.UpSampling2D(2))

​

decoder.add(layers.Conv2D(16, 3, strides=1, padding='same', activation='relu'))

decoder.add(layers.UpSampling2D(2))

​

decoder.add(layers.Conv2D(3, 3, strides=1, padding='same', activation='relu'))

decoder.add(layers.UpSampling2D(2))

​

decoder.summary()

​

// Autoencoder

​

conv_autoencoder = Model(inputs=encoder.input, outputs=decoder(encoder.outputs))

​

conv_autoencoder.compile(optimizer='adam', loss=losses.mean_squared_error)

​

history = conv_autoencoder.fit(x_train, x_train, batch_size=64, epochs=40, validation_data=(x_test, x_test))