-
Notifications
You must be signed in to change notification settings - Fork 0
/
Copy pathmodels.py
74 lines (59 loc) · 2.43 KB
/
models.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
from tensorflow.keras import Model
from layers import *
def make_discriminator():
inp = Input((None, None, 3))
h0 = Conv2D(32, (4, 4), bias_initializer=None, kernel_initializer=tf.truncated_normal_initializer(stddev=0.02), strides=2)(inp)
h0 = LeakyReLU()(h0)
h1 = Conv2D(64, (4, 4), bias_initializer=None, kernel_initializer=tf.truncated_normal_initializer(stddev=0.02), strides=2)(h0)
h1 = BatchNormalization()(h1)
h1 = LeakyReLU()(h1)
h2 = Conv2D(128, (4, 4), bias_initializer=None, kernel_initializer=tf.truncated_normal_initializer(stddev=0.02), strides=2)(h1)
h2 = BatchNormalization()(h2)
h2 = LeakyReLU()(h2)
h3 = Conv2D(256, (4, 4), bias_initializer=None, kernel_initializer=tf.truncated_normal_initializer(stddev=0.02), strides=2)(h2)
h3 = BatchNormalization()(h3)
h3 = LeakyReLU()(h3)
h4 = Conv2D(1, (3, 3), bias_initializer=None, kernel_initializer=tf.truncated_normal_initializer(stddev=0.02))(h3)
m = Model(inputs=[inp], outputs=[h4])
return m
def make_conv():
inp = Input((None, None, 3))
x = Conv2D(32, (4, 4), bias_initializer=None, kernel_initializer=tf.truncated_normal_initializer(stddev=0.02), strides=2)(inp)
x = InstanceNormalization()(x)
x = ReLU()(x)
x = Conv2D(64, (4, 4), bias_initializer=None, kernel_initializer=tf.truncated_normal_initializer(stddev=0.02), strides=2)(x)
x = InstanceNormalization()(x)
x = ReLU()(x)
return Model(inputs=[inp], outputs=[x])
def make_deconv():
inp = Input((None, None, 64))
x = ResizeConv2D(64, (4, 4), pads=(3, 3))(inp)
x = InstanceNormalization()(x)
x = ReLU()(x)
x = ResizeConv2D(32, (4, 4), pads=(3, 3))(x)
x = InstanceNormalization()(x)
x = ReLU()(x)
x = Conv2D(3, (2, 2))(x)
x = Activation('tanh')(x)
return Model(inputs=[inp], outputs=[x])
def make_slide(channels):
x = Input((None, None, channels))
r1 = Residual(dim=channels)(x)
r2 = Residual(dim=channels)(r1)
r3 = Residual(dim=channels)(r2)
r4 = Residual(dim=channels)(r3)
r5 = Residual(dim=channels)(r4)
r6 = Residual(dim=channels)(r5)
r7 = Residual(dim=channels)(r6)
model = Model(inputs=[x], outputs=[r7])
return model
def make_generator_resnet():
inp = Input((None, None, 3))
conv = make_conv()
deconv = make_deconv()
x = conv(inp)
res = make_slide(64)
res_out = res(x)
out = deconv(res_out)
m = Model(inputs=[inp], outputs=[out])
return m