-
Notifications
You must be signed in to change notification settings - Fork 10
/
Copy pathgenerator.py
164 lines (145 loc) · 8.51 KB
/
generator.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
from keras.models import Input, Model
from keras.layers import Dense, Reshape, Activation, Conv2D, Deconv2D
from keras.layers import BatchNormalization, Add, Embedding, Concatenate
import numpy as np
import keras.backend as K
from gan.layer_utils import glorot_init, resblock, dcblock
from gan.conditional_layers import ConditionalConv11, DecorelationNormalization, ConditionalCenterScale, CenterScale, FactorizedConv11
from gan.spectral_normalized_layers import SNConv2D, SNConditionalConv11, SNDense, SNEmbeding, SNFactorizedConv11
from functools import partial
def create_norm(norm, after_norm, cls=None, number_of_classes=None, filters_emb = 10,
uncoditional_conv_layer=Conv2D, conditional_conv_layer=ConditionalConv11,
factor_conv_layer=FactorizedConv11):
assert norm in ['n', 'b', 'd', 'dr']
assert after_norm in ['ucs', 'ccs', 'uccs', 'uconv', 'fconv', 'ufconv', 'cconv', 'ucconv', 'ccsuconv', 'n']
if norm == 'n':
norm_layer = lambda axis, name: (lambda inp: inp)
elif norm == 'b':
norm_layer = lambda axis, name: BatchNormalization(axis=axis, center=False, scale=False, name=name)
elif norm == 'd':
norm_layer = lambda axis, name: DecorelationNormalization(name=name)#, decomposition='zca')
elif norm == 'dr':
norm_layer = lambda axis, name: DecorelationNormalization(name=name, renorm=True)
if after_norm == 'ccs':
after_norm_layer = lambda axis, name: lambda x: ConditionalCenterScale(number_of_classes=number_of_classes,
axis=axis, name=name)([x, cls])
elif after_norm == 'ucs':
after_norm_layer = lambda axis, name: lambda x: CenterScale(axis=axis, name=name)(x)
elif after_norm == 'uccs':
def after_norm_layer(axis, name):
def f(x):
c = ConditionalCenterScale(number_of_classes=number_of_classes, axis=axis, name=name + '_c')([x, cls])
u = CenterScale(axis=axis, name=name + '_u')(x)
out = Add(name=name + '_a')([c, u])
return out
return f
elif after_norm == 'cconv':
after_norm_layer = lambda axis, name: lambda x: conditional_conv_layer(filters=K.int_shape(x)[axis],
number_of_classes=number_of_classes,
name=name)([x, cls])
elif after_norm == 'fconv':
after_norm_layer = lambda axis, name: lambda x: factor_conv_layer(number_of_classes=number_of_classes,
name=name + '_c', filters=K.int_shape(x)[axis],
filters_emb=filters_emb, use_bias=False)([x, cls])
elif after_norm == 'uconv':
after_norm_layer = lambda axis, name: lambda x: uncoditional_conv_layer(kernel_size=(1, 1),
filters=K.int_shape(x)[axis], name=name)(x)
elif after_norm == 'ucconv':
def after_norm_layer(axis, name):
def f(x):
c = conditional_conv_layer(number_of_classes=number_of_classes, name=name + '_c',
filters=K.int_shape(x)[axis])([x, cls])
u = uncoditional_conv_layer(kernel_size=(1, 1), filters=K.int_shape(x)[axis], name=name + '_u')(x)
out = Add(name=name + '_a')([c, u])
return out
return f
elif after_norm == 'ccsuconv':
def after_norm_layer(axis, name):
def f(x):
c = ConditionalCenterScale(number_of_classes=number_of_classes, axis=axis, name=name + '_c')([x, cls])
u = uncoditional_conv_layer(kernel_size=(1, 1), filters=K.int_shape(x)[axis], name=name + '_u')(x)
out = Add(name=name + '_a')([c, u])
return out
return f
elif after_norm == 'ufconv':
def after_norm_layer(axis, name):
def f(x):
c = factor_conv_layer(number_of_classes=number_of_classes, name=name + '_c',
filters=K.int_shape(x)[axis], filters_emb=filters_emb,
use_bias=False)([x, cls])
u = uncoditional_conv_layer(kernel_size=(1, 1), filters=K.int_shape(x)[axis], name=name + '_u')(x)
out = Add(name=name + '_a')([c, u])
return out
return f
elif after_norm == 'n':
after_norm_layer = lambda axis, name: lambda x: x
def result_norm(axis, name):
def stack(inp):
out = inp
out = norm_layer(axis=axis, name=name + '_npart')(out)
out = after_norm_layer(axis=axis, name=name + '_repart')(out)
return out
return stack
return result_norm
def make_generator(input_noise_shape=(128,), output_channels=3, input_cls_shape=(1, ),
block_sizes=(128, 128, 128), resamples=("UP", "UP", "UP"),
first_block_shape=(4, 4, 128), number_of_classes=10, concat_cls=False,
block_norm='u', block_after_norm='cs', filters_emb = 10,
last_norm='u', last_after_norm='cs', gan_type=None, arch='res',
spectral=False, fully_diff_spectral=False, spectral_iterations=1, conv_singular=True,):
assert arch in ['res', 'dcgan']
inp = Input(input_noise_shape)
cls = Input(input_cls_shape, dtype='int32')
if spectral:
conv_layer = partial(SNConv2D, conv_singular=conv_singular,
fully_diff_spectral=fully_diff_spectral, spectral_iterations=spectral_iterations)
cond_conv_layer = partial(SNConditionalConv11,
fully_diff_spectral=fully_diff_spectral, spectral_iterations=spectral_iterations)
dence_layer = partial(SNDense,
fully_diff_spectral=fully_diff_spectral, spectral_iterations=spectral_iterations)
emb_layer = partial(SNEmbeding, fully_diff_spectral=fully_diff_spectral, spectral_iterations=spectral_iterations)
factor_conv_layer = partial(SNFactorizedConv11,
fully_diff_spectral=fully_diff_spectral, spectral_iterations=spectral_iterations)
else:
conv_layer = Conv2D
cond_conv_layer = ConditionalConv11
dence_layer = Dense
emb_layer = Embedding
factor_conv_layer = FactorizedConv11
if concat_cls:
y = emb_layer(input_dim=number_of_classes, output_dim=first_block_shape[-1])(cls)
y = Reshape((first_block_shape[-1], ))(y)
y = Concatenate(axis=-1)([y, inp])
else:
y = inp
y = dence_layer(units=np.prod(first_block_shape), kernel_initializer=glorot_init)(y)
y = Reshape(first_block_shape)(y)
block_norm_layer = create_norm(block_norm, block_after_norm, cls=cls,
number_of_classes=number_of_classes, filters_emb=filters_emb,
uncoditional_conv_layer=conv_layer, conditional_conv_layer=cond_conv_layer,
factor_conv_layer=factor_conv_layer)
last_norm_layer = create_norm(last_norm, last_after_norm, cls=cls,
number_of_classes=number_of_classes, filters_emb=filters_emb,
uncoditional_conv_layer=conv_layer, conditional_conv_layer=cond_conv_layer,
factor_conv_layer=factor_conv_layer)
i = 0
for block_size, resample in zip(block_sizes, resamples):
if arch == 'res':
y = resblock(y, kernel_size=(3, 3), resample=resample,
nfilters=block_size, name='Generator.' + str(i),
norm=block_norm_layer, is_first=False, conv_layer=conv_layer)
else:
# TODO: SN DECONV
y = dcblock(y, kernel_size=(4, 4), resample=resample,
nfilters=block_size, name='Generator.' + str(i),
norm=block_norm_layer, is_first=False, conv_layer=Deconv2D)
i += 1
y = last_norm_layer(axis=-1, name='Generator.BN.Final')(y)
y = Activation('relu')(y)
output = conv_layer(filters=output_channels, kernel_size=(3, 3), name='Generator.Final',
kernel_initializer=glorot_init, use_bias=True, padding='same')(y)
output = Activation('tanh')(output)
if gan_type is None:
return Model(inputs=[inp], outputs=output)
else:
return Model(inputs=[inp, cls], outputs=output)