-
Notifications
You must be signed in to change notification settings - Fork 8
/
architecture.py
210 lines (181 loc) · 7.33 KB
/
architecture.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
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
from __future__ import absolute_import
from __future__ import division
from __future__ import print_function
import tensorflow as tf
import numpy as np
import tensorflow.contrib.slim as slim
from tensorflow.contrib.layers import layer_norm, instance_norm
batch_norm_params = {
"decay": 0.995,
"epsilon": 0.001,
"updates_collections": None,
"variables_collections": [tf.GraphKeys.TRAINABLE_VARIABLES],
}
gf_dim = 64
def leaky_relu(x):
return tf.maximum(0.2 * x, x)
def upscale2d(x, factor=2):
assert isinstance(factor, int) and factor >= 1
if factor == 1:
return x
with tf.variable_scope("Upscale2d"):
s = x.shape
x = tf.reshape(x, [-1, s[1], 1, s[2], 1, s[3]])
x = tf.tile(x, [1, 1, factor, 1, factor, 1])
x = tf.reshape(x, [-1, s[1] * factor, s[2] * factor, s[3]])
return x
def padding(x, pad, pad_type="reflect"):
if pad_type == "zero":
return tf.pad(x, [[0, 0], [pad, pad], [pad, pad], [0, 0]])
if pad_type == "reflect":
return tf.pad(x, [[0, 0], [pad, pad], [pad, pad], [0, 0,]], mode="REFLECT")
else:
raise ValueError("Unknown pad type: {}".format(pad_type))
def conv(x, *args, pad=1, **kwargs):
with slim.arg_scope([slim.conv2d], padding="VALID"):
x = padding(x, pad)
return slim.conv2d(x, *args, **kwargs)
def deconv(x, *args, pad=1, **kwargs):
with slim.arg_scope([slim.conv2d_transpose], padding="SAME"):
x = padding(x, pad)
return slim.conv2d(x, *args, **kwargs)
def generator(
images,
targets,
keep_prob=1.0,
phase_train=True,
weight_decay=0.0,
reuse=None,
scope="Generator",
):
with slim.arg_scope(
[slim.conv2d, slim.conv2d_transpose, slim.fully_connected],
activation_fn=tf.nn.relu,
normalizer_fn=instance_norm,
normalizer_params=None,
weights_initializer=tf.contrib.layers.variance_scaling_initializer(),
weights_regularizer=slim.l2_regularizer(weight_decay),
):
with tf.variable_scope(scope, [images], reuse=reuse):
with slim.arg_scope(
[slim.dropout, slim.batch_norm], is_training=phase_train
):
num_channels = net.shape[-1]
net = tf.concat([images, targets], axis=-1)
print(
"{} input shape: ".format(scope), [dim.value for dim in net.shape],
)
k = 64
net = conv(net, k, kernel_size=7, stride=1, pad=3, scope="conv0")
print("conv0 shape: ", [dim.value for dim in net.shape])
net = conv(net, 2 * k, kernel_size=4, stride=2, scope="conv1")
print("conv1 shape: ", [dim.value for dim in net.shape])
net = conv(net, 4 * k, kernel_size=4, stride=2, scope="conv2")
print("conv2 shape: ", [dim.value for dim in net.shape])
for i in range(3):
net_ = conv(net, 4 * k, kernel_size=3, scope="res{}_0".format(i))
net += conv(
net_,
4 * k,
3,
activation_fn=None,
biases_initializer=None,
scope="res{}_1".format(i),
)
print(
"res{} shape:".format(i), [dim.value for dim in net.shape],
)
encoded = tf.identity(net, name="encoded")
with slim.arg_scope(
[slim.conv2d, slim.conv2d_transpose, slim.fully_connected],
activation_fn=tf.nn.relu,
normalizer_fn=layer_norm,
normalizer_params=None,
weights_initializer=tf.contrib.layers.variance_scaling_initializer(),
weights_regularizer=slim.l2_regularizer(weight_decay),
):
with tf.variable_scope(scope, [encoded], reuse=reuse):
with slim.arg_scope(
[slim.dropout, slim.batch_norm], is_training=phase_train
):
with slim.arg_scope(
[slim.fully_connected],
normalizer_fn=layer_norm,
normalizer_params=None,
):
net = upscale2d(encoded, 2)
net = conv(net, 2 * k, 5, pad=2, scope="deconv1_1")
print("deconv1 shape:", [dim.value for dim in net.shape])
net = upscale2d(net, 2)
net = conv(net, k, 5, pad=2, scope="deconv2_1")
print("deconv2 shape:", [dim.value for dim in net.shape])
net = conv(
net,
num_channels,
7,
pad=3,
activation_fn=None,
normalizer_fn=None,
scope="conv_img",
)
net = tf.nn.tanh(net, name="output")
print("output:", [dim.value for dim in net.shape])
perturb = tf.clip_by_value(net, -1.0, 1.0)
output = (
2 * tf.clip_by_value(perturb + (images + 1.0) / 2.0, 0, 1) - 1
)
return net, output
def normal_discriminator(
images,
keep_prob=1.0,
phase_train=True,
weight_decay=0.0,
reuse=None,
scope="Discriminator",
):
with slim.arg_scope(
[slim.conv2d, slim.fully_connected],
weights_regularizer=slim.l2_regularizer(weight_decay),
activation_fn=leaky_relu,
normalizer_fn=slim.batch_norm,
normalizer_params=batch_norm_params,
):
with tf.variable_scope(scope, [images], reuse=reuse):
with slim.arg_scope(
[slim.batch_norm, slim.dropout], is_training=phase_train
):
print(
"{} input shape:".format(scope),
[dim.value for dim in images.shape],
)
net = conv(
images,
32,
kernel_size=4,
stride=2,
scope="conv1",
activation_fn=None,
)
print("module_1 shape:", [dim.value for dim in net.shape])
net = conv(
net, 64, kernel_size=4, stride=2, scope="conv2", activation_fn=None,
)
print("module_2 shape:", [dim.value for dim in net.shape])
net = conv(net, 128, kernel_size=4, stride=2, scope="conv3")
print("module_3 shape:", [dim.value for dim in net.shape])
net = conv(net, 256, kernel_size=4, stride=2, scope="conv4")
print("module_4 shape:", [dim.value for dim in net.shape])
net = conv(net, 512, kernel_size=4, stride=2, scope="conv5")
print("module_5 shape:", [dim.value for dim in net.shape])
net = slim.conv2d(
net,
1,
1,
activation_fn=None,
normalizer_fn=None,
scope="patch_logits",
)
print("patch:", [dim.value for dim in net.shape])
net = tf.reshape(net, [-1, 1])
print("disc shape: ", [dim.value for dim in net.shape])
return net