python - Split autoencoder on encoder and decoder keras

Question

I am trying to create an autoencoder for:

Train the model
Split encoder and decoder
Visualise compressed data (encoder)
Use arbitrary compressed data to get the output (decoder)

from keras.layers import Input, Dense, Conv2D, MaxPooling2D, UpSampling2D
from keras.models import Model
from keras import backend as K
from keras.datasets import mnist
import numpy as np

(x_train, _), (x_test, _) = mnist.load_data()

x_train = x_train.astype('float32') / 255.
x_train = x_train[:100,:,:,]
x_test = x_test.astype('float32') / 255.
x_test = x_train
x_train = np.reshape(x_train, (len(x_train), 28, 28, 1))  # adapt this if using `channels_first` image data format
x_test = np.reshape(x_test, (len(x_test), 28, 28, 1))  # adapt this if using `channels_first` image data format
 input_img = Input(shape=(28, 28, 1))  # adapt this if using `channels_first` image data format

x = Conv2D(32, (3, 3), activation='relu', padding='same')(input_img)
x = MaxPooling2D((2, 2), padding='same')(x)
x = Conv2D(32, (3, 3), activation='relu', padding='same')(x)
encoded = MaxPooling2D((2, 2), padding='same')(x)

# at this point the representation is (7, 7, 32)

decoder = Conv2D(32, (3, 3), activation='relu', padding='same')(encoded)
x = UpSampling2D((2, 2))(decoder)
x = Conv2D(32, (3, 3), activation='relu', padding='same')(x)
x = UpSampling2D((2, 2))(x)
decoded = Conv2D(1, (3, 3), activation='sigmoid', padding='same')(x)

autoencoder = Model(input_img, decoded(encoded(input_img)))
autoencoder.compile(optimizer='adadelta', loss='binary_crossentropy')

autoencoder.fit(x_train, x_train,
                epochs=10,
                batch_size=128,
                shuffle=True,
                validation_data=(x_test, x_test),
                #callbacks=[TensorBoard(log_dir='/tmp/tb', histogram_freq=0, write_graph=False)]
               )

How to split train it and split with the trained weights?

@desertnaut I am so sorry! Now I edited the question. The main difficulty here is here autoencoder = Model(input_img, decoded(encoded(input_img))) — ziji
I think the OP wants to feed an external input to the decoder and observe the output. I'm not sure how to do this with keras, but with tensorflow you could use tf.placeholder_with_default() which passes through the input when the placeholder is not fed. — Achintha Ihalage

Daniel Möller Daniel Möller · Accepted Answer · 2019-02-28T17:30:41

Make encoder:

input_img = Input(shape=(28, 28, 1))

x = Conv2D(32, (3, 3), activation='relu', padding='same')(input_img)
x = MaxPooling2D((2, 2), padding='same')(x)
x = Conv2D(32, (3, 3), activation='relu', padding='same')(x)
encoded = MaxPooling2D((2, 2), padding='same')(x)

encoder = Model(input_img, encoded)

Make decoder:

decoder_input= Input(shape_equal_to_encoder_output_shape)

decoder = Conv2D(32, (3, 3), activation='relu', padding='same')(decoder_input)
x = UpSampling2D((2, 2))(decoder)
x = Conv2D(32, (3, 3), activation='relu', padding='same')(x)
x = UpSampling2D((2, 2))(x)
decoded = Conv2D(1, (3, 3), activation='sigmoid', padding='same')(x)

decoder = Model(decoder_input, decoded)

Make autoencoder:

auto_input = Input(shape=(28,28,1))
encoded = encoder(auto_input)
decoded = decoder(encoded)

auto_encoder = Model(auto_input, decoded)

Now you can use any of them any way you want to.

train the autoencoder
use the encoder and decoder

python - Split autoencoder on encoder and decoder keras

1 Answers