-
Notifications
You must be signed in to change notification settings - Fork 0
/
Copy pathQ3_gan.py
169 lines (134 loc) · 5.81 KB
/
Q3_gan.py
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
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
from keras.preprocessing.image import ImageDataGenerator, array_to_img, img_to_array, load_img
from keras.models import Sequential, Model
from keras.layers import Conv2D, MaxPooling2D, UpSampling2D, Conv2DTranspose, Reshape
from keras.layers import Activation, Dropout, Flatten, Dense, LeakyReLU, Input, Lambda, BatchNormalization
from keras.callbacks import EarlyStopping
from keras.datasets import mnist
from keras.losses import binary_crossentropy
from keras import optimizers, regularizers
import numpy as np
from sklearn import metrics
from keras.utils import to_categorical
from sklearn.cluster import KMeans
import joblib
from scipy.io import loadmat
class MyEarlyStopping(EarlyStopping):
def __init__(self, *args, **kw):
super().__init__(*args, **kw)
self.baseline_attained = False
def on_epoch_end(self, epoch, logs=None):
if not self.baseline_attained:
current = self.get_monitor_value(logs)
if current is None:
return
if self.monitor_op(current, self.baseline):
if self.verbose > 0:
print('Baseline attained.')
self.baseline_attained = True
self.model.stop_training = True
else:
return
super(MyEarlyStopping, self).on_epoch_end(epoch, logs)
img_rows = 28
img_cols = 20
########################## Load frey ##########################
file = loadmat("./frey_rawface.mat", squeeze_me=True, struct_as_record=False)
file = file["ff"].T.reshape((-1, img_rows, img_cols))
np.random.seed(42)
original_dim = 28 * 20
x_train = file
x_train = x_train.astype('float32') / 255.
# x_train = x_train.reshape((len(x_train), original_dim))
# x_test = x_test.reshape((len(x_test), original_dim))
x_train = x_train.reshape(-1, img_rows, img_cols, 1).astype(np.float32)
print(x_train.shape)
########################## Create Discriminator ##########################
discriminator = Sequential()
depth = 64
dropout = 0.4
# In: 28 x 28 x 1, depth = 1
# Out: 14 x 14 x 1, depth=64
input_shape = (img_rows, img_cols, 1)
discriminator.add(Conv2D(depth*1, 5, strides=2, input_shape=input_shape,\
padding='same', activation=LeakyReLU(alpha=0.2)))
discriminator.add(Dropout(dropout))
discriminator.add(Conv2D(depth*2, 5, strides=2, padding='same',\
activation=LeakyReLU(alpha=0.2)))
discriminator.add(Dropout(dropout))
discriminator.add(Conv2D(depth*4, 5, strides=2, padding='same',\
activation=LeakyReLU(alpha=0.2)))
discriminator.add(Dropout(dropout))
discriminator.add(Conv2D(depth*8, 5, strides=1, padding='same',\
activation=LeakyReLU(alpha=0.2)))
discriminator.add(Dropout(dropout))
# Out: 1-dim probability
discriminator.add(Flatten())
discriminator.add(Dense(1))
discriminator.add(Activation('sigmoid'))
print(discriminator.summary())
discriminator.compile(loss='binary_crossentropy', optimizer=optimizers.RMSprop(lr=0.0008, clipvalue=1.0, decay=6e-8), metrics=['accuracy'])
########################## Create Generator ##########################
generator = Sequential()
dropout = 0.4
depth = 254
dim = 7
generator.add(Dense(7*5*depth, input_dim=100))
generator.add(BatchNormalization(momentum=0.9))
generator.add(Activation('relu'))
generator.add(Reshape((7, 5, depth)))
generator.add(Dropout(dropout))
generator.add(UpSampling2D())
generator.add(Conv2DTranspose(int(depth/2), 5, padding='same'))
generator.add(BatchNormalization(momentum=0.9))
generator.add(Activation('relu'))
generator.add(UpSampling2D())
generator.add(Conv2DTranspose(int(depth/4), 5, padding='same'))
generator.add(BatchNormalization(momentum=0.9))
generator.add(Activation('relu'))
generator.add(Conv2DTranspose(int(depth/8), 5, padding='same'))
generator.add(BatchNormalization(momentum=0.9))
generator.add(Activation('relu'))
generator.add(Conv2DTranspose(1, 5, padding='same'))
generator.add(Activation('sigmoid'))
print(generator.summary())
generator.compile(loss='binary_crossentropy', optimizer=optimizers.RMSprop(lr=0.0004, clipvalue=1.0, decay=3e-8),metrics=['accuracy'])
gan = Sequential()
gan.add(generator)
gan.add(discriminator)
gan.compile(loss='binary_crossentropy', optimizer=optimizers.RMSprop(lr=0.0004, clipvalue=1.0, decay=3e-8),metrics=['accuracy'])
print(gan.summary())
train_steps=2000
batch_size=256
import matplotlib.pyplot as plt
########################## Train GAN and save faces ##########################
for i in range(train_steps):
print(x_train.shape)
images_train = x_train[np.random.randint(0,x_train.shape[0], size=batch_size)]
noise = np.random.uniform(-1.0, 1.0, size=[batch_size, 100])
images_fake = generator.predict(noise)
print(images_train.shape, images_fake.shape)
x = np.concatenate((images_train, images_fake))
y = np.ones([2*batch_size, 1])
y[batch_size:, :] = 0
d_loss = discriminator.train_on_batch(x, y)
y = np.ones([batch_size, 1])
noise = np.random.uniform(-1.0, 1.0, size=[batch_size, 100])
a_loss = gan.train_on_batch(noise, y)
print("%d [discriminator loss: %f, acc: %f] [gan loss: %f, acc: %f]" % (i, d_loss[0], d_loss[1], a_loss[0], a_loss[1]))
if(i%50 == 0):
filename = "frey_%d.png" % i
noise = np.random.uniform(-1.0, 1.0, size=[16, 100])
images = generator.predict(noise)
plt.figure(figsize=(28,20))
for i in range(images.shape[0]):
print(i, images.shape[0])
plt.subplot(4, 4, i+1)
image = images[i, :, :, :]
image = np.reshape(image, [img_rows, img_cols])
plt.imshow(image, cmap='gray')
plt.axis('off')
plt.tight_layout()
plt.savefig(filename)
plt.show()
plt.close('all')
gan.save_weights('gan_model.h5')