-
Notifications
You must be signed in to change notification settings - Fork 0
/
models.py
139 lines (97 loc) · 6.58 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
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
from typing import Tuple, List, Text, Dict, Any, Iterator
from keras.models import Model
from keras.layers import Input
from keras.layers.core import Activation, Dropout, Flatten, Dense
from keras.layers.merge import Concatenate
from keras.layers.convolutional import Conv2D, Conv2DTranspose
from keras.layers.normalization import BatchNormalization
from keras.layers.advanced_activations import LeakyReLU
from keras.utils import plot_model
def create_unet(in_shape: Tuple[int,int,int], out_shape: Tuple[int,int,int], filters: int) -> Model:
'''
reference models
* https://github.com/phillipi/pix2pix/blob/master/models.lua#L47
* https://github.com/tdeboissiere/DeepLearningImplementations/blob/master/pix2pix/src/model/models.py#L317
'''
input_tensor = Input(shape=in_shape) # type: Input
output_ch = out_shape[2]
# enc
x = Conv2D( filters*1, kernel_size=(4, 4), strides=(2, 2), padding="same")( input_tensor ) ; e1 = x
x = BatchNormalization()( Conv2D( filters*2, kernel_size=(4, 4), strides=(2, 2), padding="same")( LeakyReLU(0.2)(x) ) ); e2 = x
x = BatchNormalization()( Conv2D( filters*4, kernel_size=(4, 4), strides=(2, 2), padding="same")( LeakyReLU(0.2)(x) ) ); e3 = x
x = BatchNormalization()( Conv2D( filters*8, kernel_size=(4, 4), strides=(2, 2), padding="same")( LeakyReLU(0.2)(x) ) ); e4 = x
x = BatchNormalization()( Conv2D( filters*8, kernel_size=(4, 4), strides=(2, 2), padding="same")( LeakyReLU(0.2)(x) ) ); e5 = x
x = BatchNormalization()( Conv2D( filters*8, kernel_size=(4, 4), strides=(2, 2), padding="same")( LeakyReLU(0.2)(x) ) ); e6 = x
x = BatchNormalization()( Conv2D( filters*8, kernel_size=(4, 4), strides=(2, 2), padding="same")( LeakyReLU(0.2)(x) ) ); e7 = x
x = Conv2D( filters*8, kernel_size=(4, 4), strides=(2, 2), padding="same")( LeakyReLU(0.2)(x) ) ; e8 = x
# dec
x = BatchNormalization()( Conv2DTranspose(filters*8, kernel_size=(4, 4), strides=(2, 2), padding="same")( Activation("relu")(x) ) ); x = Concatenate()([Dropout(0.5)(x), e7])
x = BatchNormalization()( Conv2DTranspose(filters*8, kernel_size=(4, 4), strides=(2, 2), padding="same")( Activation("relu")(x) ) ); x = Concatenate()([Dropout(0.5)(x), e6])
x = BatchNormalization()( Conv2DTranspose(filters*8, kernel_size=(4, 4), strides=(2, 2), padding="same")( Activation("relu")(x) ) ); x = Concatenate()([Dropout(0.5)(x), e5])
x = BatchNormalization()( Conv2DTranspose(filters*8, kernel_size=(4, 4), strides=(2, 2), padding="same")( Activation("relu")(x) ) ); x = Concatenate()([x, e4])
x = BatchNormalization()( Conv2DTranspose(filters*4, kernel_size=(4, 4), strides=(2, 2), padding="same")( Activation("relu")(x) ) ); x = Concatenate()([x, e3])
x = BatchNormalization()( Conv2DTranspose(filters*2, kernel_size=(4, 4), strides=(2, 2), padding="same")( Activation("relu")(x) ) ); x = Concatenate()([x, e2])
x = BatchNormalization()( Conv2DTranspose(filters*1, kernel_size=(4, 4), strides=(2, 2), padding="same")( Activation("relu")(x) ) ); x = Concatenate()([x, e1])
x = Conv2DTranspose(output_ch, kernel_size=(4, 4), strides=(2, 2), padding="same")( Activation("relu")(x) )
x = Activation("tanh")(x)
unet = Model(inputs=[input_tensor], outputs=[x])
return unet
def create_discriminator_patch(in_shape: Tuple[int,int,int], filters: int) -> Model:
'''
PatchGAN
reference models
* https://github.com/phillipi/pix2pix/blob/b479b6b7d37f9d7e87dce7f5e627dc3bb7b4a117/models.lua#L180
* https://github.com/makora9143/pix2pix-keras-tensorflow/blob/4b7d2192607448659ba7b2c0b638d395dcd23ef4/model.py#L13
* https://github.com/tdeboissiere/DeepLearningImplementations/blob/master/pix2pix/src/model/models.py#L201
patch_shape = (70, 70)
if generator: (256, 256, 3) -> (256, 256, 3)
then inshape == (70, 70, 3+3)
'''
input_tensor = Input(shape=in_shape) # type: Input
x = LeakyReLU(0.2)( Conv2D(filters*1, kernel_size=(4, 4), strides=(2, 2), padding="same")( input_tensor ) )
x = LeakyReLU(0.2)( BatchNormalization()( Conv2D(filters*2, kernel_size=(4, 4), strides=(2, 2), padding="same")(x) ) )
x = LeakyReLU(0.2)( BatchNormalization()( Conv2D(filters*4, kernel_size=(4, 4), strides=(2, 2), padding="same")(x) ) )
x = LeakyReLU(0.2)( BatchNormalization()( Conv2D(filters*8, kernel_size=(4, 4), strides=(2, 2), padding="same")(x) ) )
x = Activation("sigmoid")( Conv2D(filters*1, kernel_size=(4, 4), strides=(2, 2), padding="same")(x) )
x = Flatten()(x)#; x_flat = x # to 1 dim array to use minibatch discrimination
#x = Activation("softmax")( Dense(2)(x) )
disc = Model(input=[input_tensor], output=[x])#, x_flat])
return disc
def create_discriminator_image(in_shape: Tuple[int,int,int], filters: int):
'''
ImageGAN
reference models
* https://github.com/costapt/vess2ret/blob/master/models.py#L517
'''
return create_discriminator_patch(in_shape, filters)
def pix2pix():
unet = create_unet((256, 256, 3), (256, 256, 3), 128)
disc = create_discriminator_image((256, 256, 6), 64)
a_real = Input(shape=(a_ch, 512, 512))
b_real = Input(shape=(b_ch, 512, 512))
b_fake = unet(a_real)
disc(Concatenate()([a_real, b_real]))
disc(Concatenate()([a_real, b_fake]))
if __name__ == '__main__':
unet = create_unet((256, 256, 3), (256, 256, 3), 128)
unet.summary()
plot_model(unet, to_file='unet.png', show_shapes=True, show_layer_names=True)
disc = create_discriminator_image((256, 256, 6), 64)
disc.summary()
plot_model(disc, to_file='disc.png', show_shapes=True, show_layer_names=True)
exit()
def ___create_discriminator(in_shape: Tuple[int,int,int], filters: int):
'''
working on prog
reference models
* https://github.com/tdeboissiere/DeepLearningImplementations/blob/master/pix2pix/src/model/models.py#L201
'''
input_tensors = [Input(shape=(70, 70, 3)) for i in range(nb_patch)]
patches = [ create_discriminator_patch(input_tensor, filters) for input_tensor in input_tensors]
xs = [x for (x, _) in patches]
#x_flats = [x_flat for (_, x_flat) in patches]
x = Activation("softmax")( Dense(2)( Concatenate()(xs) ) )
#x_mbd = Activation("softmax")( Dense(2)( Concatenate()(x_flats) ) )
# todo: minibatch discrimination - https://github.com/tdeboissiere/DeepLearningImplementations/blob/master/pix2pix/src/model/models.py#L250
discriminator = Model(inputs=input_tensors, outputs=[x])
return discriminator