16.卷积自编码器

程序说明

时间：2016年11月23日

说明：一个卷积自编码器的例程。

数据集：MNIST

原博客地址：Building Autoencoders in Keras

1.加载keras模块

from keras.layers import Input, Dense, Convolution2D, MaxPooling2D, UpSampling2D
from keras.models import Model
from keras.datasets import mnist
import numpy as np
import matplotlib.pyplot as plt
%matplotlib inline

Using TensorFlow backend.

2.Convolutional autoencoder

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

x = Convolution2D(16, 3, 3, activation='relu', border_mode='same')(input_img)
x = MaxPooling2D((2, 2), border_mode='same')(x)
x = Convolution2D(8, 3, 3, activation='relu', border_mode='same')(x)
x = MaxPooling2D((2, 2), border_mode='same')(x)
x = Convolution2D(8, 3, 3, activation='relu', border_mode='same')(x)
encoded = MaxPooling2D((2, 2), border_mode='same')(x)

# at this point the representation is (8, 4, 4) i.e. 128-dimensional

x = Convolution2D(8, 3, 3, activation='relu', border_mode='same')(encoded)
x = UpSampling2D((2, 2))(x)
x = Convolution2D(8, 3, 3, activation='relu', border_mode='same')(x)
x = UpSampling2D((2, 2))(x)
x = Convolution2D(16, 3, 3, activation='relu')(x)
x = UpSampling2D((2, 2))(x)
decoded = Convolution2D(1, 3, 3, activation='sigmoid', border_mode='same')(x)

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

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

x_train = x_train.astype('float32') / 255.
x_test = x_test.astype('float32') / 255.
x_train = np.reshape(x_train, (len(x_train), 28, 28, 1))
x_test = np.reshape(x_test, (len(x_test), 28, 28, 1))

autoencoder.fit(x_train, x_train,
                nb_epoch=50,
                batch_size=128,
                shuffle=True,
                validation_data=(x_test, x_test))

Train on 60000 samples, validate on 10000 samples
Epoch 1/50
60000/60000 [==============================] - 8s - loss: 0.2372 - val_loss: 0.1694
Epoch 2/50
60000/60000 [==============================] - 7s - loss: 0.1588 - val_loss: 0.1502
Epoch 3/50
60000/60000 [==============================] - 7s - loss: 0.1441 - val_loss: 0.1346
Epoch 4/50
60000/60000 [==============================] - 7s - loss: 0.1353 - val_loss: 0.1313
Epoch 5/50
60000/60000 [==============================] - 7s - loss: 0.1293 - val_loss: 0.1266
Epoch 6/50
60000/60000 [==============================] - 7s - loss: 0.1252 - val_loss: 0.1229
Epoch 7/50
60000/60000 [==============================] - 7s - loss: 0.1222 - val_loss: 0.1193
Epoch 8/50
60000/60000 [==============================] - 7s - loss: 0.1202 - val_loss: 0.1179
Epoch 9/50
60000/60000 [==============================] - 7s - loss: 0.1184 - val_loss: 0.1185
Epoch 10/50
60000/60000 [==============================] - 7s - loss: 0.1169 - val_loss: 0.1166
Epoch 11/50
60000/60000 [==============================] - 7s - loss: 0.1155 - val_loss: 0.1132
Epoch 12/50
60000/60000 [==============================] - 7s - loss: 0.1143 - val_loss: 0.1141
Epoch 13/50
60000/60000 [==============================] - 7s - loss: 0.1134 - val_loss: 0.1136
Epoch 14/50
60000/60000 [==============================] - 7s - loss: 0.1125 - val_loss: 0.1124
Epoch 15/50
60000/60000 [==============================] - 7s - loss: 0.1119 - val_loss: 0.1091
Epoch 16/50
60000/60000 [==============================] - 7s - loss: 0.1111 - val_loss: 0.1090
Epoch 17/50
60000/60000 [==============================] - 7s - loss: 0.1103 - val_loss: 0.1120
Epoch 18/50
60000/60000 [==============================] - 7s - loss: 0.1094 - val_loss: 0.1074
Epoch 19/50
60000/60000 [==============================] - 7s - loss: 0.1091 - val_loss: 0.1070
Epoch 20/50
60000/60000 [==============================] - 7s - loss: 0.1087 - val_loss: 0.1085
Epoch 21/50
60000/60000 [==============================] - 7s - loss: 0.1082 - val_loss: 0.1069
Epoch 22/50
60000/60000 [==============================] - 7s - loss: 0.1079 - val_loss: 0.1058
Epoch 23/50
60000/60000 [==============================] - 7s - loss: 0.1075 - val_loss: 0.1045
Epoch 24/50
60000/60000 [==============================] - 7s - loss: 0.1067 - val_loss: 0.1055
Epoch 25/50
60000/60000 [==============================] - 7s - loss: 0.1067 - val_loss: 0.1070
Epoch 26/50
60000/60000 [==============================] - 7s - loss: 0.1063 - val_loss: 0.1032
Epoch 27/50
60000/60000 [==============================] - 7s - loss: 0.1058 - val_loss: 0.1060
Epoch 28/50
60000/60000 [==============================] - 7s - loss: 0.1055 - val_loss: 0.1069
Epoch 29/50
60000/60000 [==============================] - 7s - loss: 0.1053 - val_loss: 0.1037
Epoch 30/50
60000/60000 [==============================] - 7s - loss: 0.1051 - val_loss: 0.1055
Epoch 31/50
60000/60000 [==============================] - 7s - loss: 0.1048 - val_loss: 0.1038
Epoch 32/50
60000/60000 [==============================] - 7s - loss: 0.1045 - val_loss: 0.1024
Epoch 33/50
60000/60000 [==============================] - 7s - loss: 0.1043 - val_loss: 0.1022
Epoch 34/50
60000/60000 [==============================] - 7s - loss: 0.1041 - val_loss: 0.1023
Epoch 35/50
60000/60000 [==============================] - 7s - loss: 0.1040 - val_loss: 0.1042
Epoch 36/50
60000/60000 [==============================] - 7s - loss: 0.1038 - val_loss: 0.1030
Epoch 37/50
60000/60000 [==============================] - 7s - loss: 0.1036 - val_loss: 0.1020
Epoch 38/50
60000/60000 [==============================] - 7s - loss: 0.1033 - val_loss: 0.1015
Epoch 39/50
60000/60000 [==============================] - 7s - loss: 0.1032 - val_loss: 0.1010
Epoch 40/50
60000/60000 [==============================] - 7s - loss: 0.1027 - val_loss: 0.1023
Epoch 41/50
60000/60000 [==============================] - 7s - loss: 0.1025 - val_loss: 0.1000
Epoch 42/50
60000/60000 [==============================] - 7s - loss: 0.1024 - val_loss: 0.1016
Epoch 43/50
60000/60000 [==============================] - 7s - loss: 0.1018 - val_loss: 0.1001
Epoch 44/50
60000/60000 [==============================] - 7s - loss: 0.1016 - val_loss: 0.1000
Epoch 45/50
60000/60000 [==============================] - 7s - loss: 0.1016 - val_loss: 0.1004
Epoch 46/50
60000/60000 [==============================] - 7s - loss: 0.1013 - val_loss: 0.0993
Epoch 47/50
60000/60000 [==============================] - 7s - loss: 0.1010 - val_loss: 0.1006
Epoch 48/50
60000/60000 [==============================] - 7s - loss: 0.1006 - val_loss: 0.1001
Epoch 49/50
60000/60000 [==============================] - 7s - loss: 0.1008 - val_loss: 0.0994
Epoch 50/50
60000/60000 [==============================] - 7s - loss: 0.1006 - val_loss: 0.0998





<keras.callbacks.History at 0x7f0244252a50>

decoded_imgs = autoencoder.predict(x_test)

n = 10
plt.figure(figsize=(20, 4))
for i in range(n):
    # display original
    ax = plt.subplot(2, n, i+1)
    plt.imshow(x_test[i].reshape(28, 28))
    plt.gray()
    ax.get_xaxis().set_visible(False)
    ax.get_yaxis().set_visible(False)

    # display reconstruction
    ax = plt.subplot(2, n, i + n+1)
    plt.imshow(decoded_imgs[i].reshape(28, 28))
    plt.gray()
    ax.get_xaxis().set_visible(False)
    ax.get_yaxis().set_visible(False)
plt.show()

autoencoder.summary()

____________________________________________________________________________________________________
Layer (type)                     Output Shape          Param #     Connected to                     
====================================================================================================
input_1 (InputLayer)             (None, 28, 28, 1)     0                                            
____________________________________________________________________________________________________
convolution2d_1 (Convolution2D)  (None, 28, 28, 16)    160         input_1[0][0]                    
____________________________________________________________________________________________________
maxpooling2d_1 (MaxPooling2D)    (None, 14, 14, 16)    0           convolution2d_1[0][0]            
____________________________________________________________________________________________________
convolution2d_2 (Convolution2D)  (None, 14, 14, 8)     1160        maxpooling2d_1[0][0]             
____________________________________________________________________________________________________
maxpooling2d_2 (MaxPooling2D)    (None, 7, 7, 8)       0           convolution2d_2[0][0]            
____________________________________________________________________________________________________
convolution2d_3 (Convolution2D)  (None, 7, 7, 8)       584         maxpooling2d_2[0][0]             
____________________________________________________________________________________________________
maxpooling2d_3 (MaxPooling2D)    (None, 4, 4, 8)       0           convolution2d_3[0][0]            
____________________________________________________________________________________________________
convolution2d_4 (Convolution2D)  (None, 4, 4, 8)       584         maxpooling2d_3[0][0]             
____________________________________________________________________________________________________
upsampling2d_1 (UpSampling2D)    (None, 8, 8, 8)       0           convolution2d_4[0][0]            
____________________________________________________________________________________________________
convolution2d_5 (Convolution2D)  (None, 8, 8, 8)       584         upsampling2d_1[0][0]             
____________________________________________________________________________________________________
upsampling2d_2 (UpSampling2D)    (None, 16, 16, 8)     0           convolution2d_5[0][0]            
____________________________________________________________________________________________________
convolution2d_6 (Convolution2D)  (None, 14, 14, 16)    1168        upsampling2d_2[0][0]             
____________________________________________________________________________________________________
upsampling2d_3 (UpSampling2D)    (None, 28, 28, 16)    0           convolution2d_6[0][0]            
____________________________________________________________________________________________________
convolution2d_7 (Convolution2D)  (None, 28, 28, 1)     145         upsampling2d_3[0][0]             
====================================================================================================
Total params: 4385
____________________________________________________________________________________________________

中间编码层编码可视化

model_extractfeatures = Model(input=autoencoder.input, output=autoencoder.get_layer('maxpooling2d_3').output)
encoded_imgs = model_extractfeatures.predict(x_test)
print(encoded_imgs.shape)
n = 10
plt.figure(figsize=(20, 8))
for i in range(n):
    ax = plt.subplot(1, n, i+1)
    plt.imshow(encoded_imgs[i].reshape(4, 4 * 8).T)
    plt.gray()
    ax.get_xaxis().set_visible(False)
    ax.get_yaxis().set_visible(False)
plt.show()

(10000, 4, 4, 8)