AutoEncorder/MNISTTEST.py at master · kakeru58/AutoEncorder · GitHub

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
from keras.layers import Input, Dense, Lambda, Conv2D, Flatten, Reshape, Conv2DTranspose, MaxPooling2D, UpSampling2D
from keras.models import Model
from keras.datasets import mnist
from keras.losses import binary_crossentropy
import numpy as np
from sklearn.model_selection import train_test_split
import matplotlib.pyplot as plt
from keras import regularizers
from keras import backend as K
import cv2
# MNISTデータを前処理する
(x_train, y_train), (x_test, y_test) = mnist.load_data()  # 使わないため_に
img_size = x_train.shape[1]
x_train, x_valid, y_train, y_valid = train_test_split(x_train, y_train, test_size=0.175)  # 訓練ようと試験用に分ける0.175は割合 過学習防止
x_train = x_train.astype('float32') / 255.  # int型を画素ちの最大255でわる．(正規化)
x_valid = x_valid.astype('float32') / 255.
x_test = x_test.astype('float32') / 255.

x_train = np.reshape(x_train, (len(x_train), img_size, img_size, 1))  # 3次元データを2次元データにlenは要素数
x_valid = np.reshape(x_valid, (len(x_valid), img_size, img_size, 1))  # np.prodで1番目からの要素の積(784)
x_test = np.reshape(x_test, (len(x_test), img_size, img_size, 1))

# 入力用の変数
x = input_img = Input(shape=(img_size, img_size, 1))  # 28*28
batch_size = 64
latent_dim = 2
epoch = 50
x = Conv2D(32, (3, 3), activation='relu', strides=1, padding='same')(x)
x = Conv2D(64, (3, 3), activation='relu', strides=1, padding='same')(x)

shape = K.int_shape(x)
x = Flatten()(x)
x = Dense(16, activation='relu')(x)
# 潜在変数
z_mean = Dense(latent_dim)(x)
z_log_var = Dense(latent_dim)(x)


def sampling(args):
    z_mean, z_log_var = args
    batch = K.shape(z_mean)[0]
    dim = K.int_shape(z_mean)[1]
    # default=random_normal has mean=0 and std=1.0
    epsilon = K.random_normal(shape=(batch, dim))
    return z_mean + K.exp(0.5 * z_log_var) * epsilon


z = Lambda(sampling)([z_mean, z_log_var])

encoder = Model(input_img, [z_mean, z_log_var, z])
encoder.summary()
# decorder
latent_inputs = Input(shape=(latent_dim,), name='z_sampling')
x = Dense(shape[1] * shape[2] * shape[3], activation='relu')(latent_inputs)
x = Reshape((shape[1], shape[2], shape[3]))(x)

x = Conv2DTranspose(64, (3, 3), activation='relu', strides=1, padding='same')(x)
x = Conv2DTranspose(32, (3, 3), activation='relu', strides=1, padding='same')(x)
decoded = Conv2DTranspose(1, (3, 3), activation='sigmoid', padding='same')(x)

decoder = Model(latent_inputs, decoded)
decoder.summary()
decoded = decoder(encoder(input_img)[2])
autoencoder = Model(input_img, decoded)  # 入力と出力

# loss関数
# Compute VAE loss
reconstruction_loss = binary_crossentropy(K.flatten(input_img),
                                          K.flatten(decoded))
reconstruction_loss *= img_size * img_size
kl_loss = 1 + z_log_var - K.square(z_mean) - K.exp(z_log_var)
kl_loss = K.sum(kl_loss, axis=-1)
kl_loss *= -0.5
vae_loss = K.mean(reconstruction_loss + kl_loss)
# AdaDeltaで最適化, loss関数はbinary_crossentropy
autoencoder.add_loss(vae_loss)
autoencoder.compile(optimizer='adam', loss='')
autoencoder.summary()
x_train = x_train[y_train == 1]
x_valid = x_valid[y_valid == 1]
# autoencoderの実行
history = autoencoder.fit(x_train,  # kerasのfitメゾット
                          epochs=epoch,
                          batch_size=batch_size,
                          validation_data=(x_valid, None))
# 画像化して確認
decoded_img = autoencoder.predict(x_test)

n = 10  # (表示個数)
plt.figure(figsize=(20, 6))  # 横インチ縦インチの大きさ
for i in range(n):
    # オリジナル
    origin_img = x_test[i].reshape(28, 28)
    ax = plt.subplot(3, n, i + 1)  # 何行何列何番目
    plt.imshow(origin_img)
    plt.gray()
    ax.get_xaxis().set_visible(False)  # 軸の値非表示
    ax.get_yaxis().set_visible(False)
    # 変換された画像
    change_img = decoded_img[i].reshape(28, 28)
    ax = plt.subplot(3, n, i + 1 + n)
    plt.imshow(change_img)
    plt.gray()
    ax.get_xaxis().set_visible(False)
    ax.get_yaxis().set_visible(False)
    # 差分
    diff_img = origin_img - change_img
    diff_img_abs = np.abs(diff_img)
    diff_img = diff_img_abs * 255
    ax = plt.subplot(3, n, i + 1 + 2 * n)
    plt.imshow(diff_img)
    plt.gray()
    ax.get_xaxis().set_visible(False)
    ax.get_yaxis().set_visible(False)
plt.show()
# 損失関数のグラフの軸ラベルを設定
plt.xlabel('time step')
plt.ylabel('loss')
# グラフ縦軸の範囲を0以上と定める
plt.ylim(0, max(np.r_[history.history['val_loss'], history.history['loss']]))
# 損失関数の時間変化を描画
val_loss, = plt.plot(history.history['val_loss'], c='#56B4E9')
loss, = plt.plot(history.history['loss'], c='#E69F00')
# グラフの凡例（はんれい）を追加
plt.legend([loss, val_loss], ['loss', 'val_loss'])
# 描画したグラフを表示
plt.show()


# https://qiita.com/fukuit/items/2f8bdbd36979fff96b07