-
Notifications
You must be signed in to change notification settings - Fork 0
/
model.py
121 lines (107 loc) · 6.15 KB
/
model.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
from keras.models import Model
from keras.layers import Conv2D, Input, MaxPooling2D, Dropout, UpSampling2D, concatenate
from kaf import KAF
def unet(
input_size=(256, 256, 3),
activation="relu",
pretrained_weights=None,
kaf_kernel="softplus",
init_fcn=None,
kaf_D=5,
):
if activation == "kaf":
activation = None
inputs = Input(input_size)
conv1 = Conv2D(64, 3, activation=activation, padding="same", kernel_initializer="he_normal")(inputs)
if activation is None:
conv1 = KAF(64, D=kaf_D, kernel=kaf_kernel, init_fcn=init_fcn, conv=True)(conv1)
conv1 = Conv2D(64, 3, activation=activation, padding="same", kernel_initializer="he_normal")(conv1)
if activation is None:
conv1 = KAF(64, D=kaf_D, kernel=kaf_kernel, init_fcn=init_fcn, conv=True)(conv1)
pool1 = MaxPooling2D(pool_size=(2, 2))(conv1)
conv2 = Conv2D(128, 3, activation=activation, padding="same", kernel_initializer="he_normal")(pool1)
if activation is None:
conv2 = KAF(128, D=kaf_D, kernel=kaf_kernel, init_fcn=init_fcn, conv=True)(conv2)
conv2 = Conv2D(128, 3, activation=activation, padding="same", kernel_initializer="he_normal")(conv2)
if activation is None:
conv2 = KAF(128, D=kaf_D, kernel=kaf_kernel, init_fcn=init_fcn, conv=True)(conv2)
pool2 = MaxPooling2D(pool_size=(2, 2))(conv2)
conv3 = Conv2D(256, 3, activation=activation, padding="same", kernel_initializer="he_normal")(pool2)
if activation is None:
conv3 = KAF(256, D=kaf_D, kernel=kaf_kernel, init_fcn=init_fcn, conv=True)(conv3)
conv3 = Conv2D(256, 3, activation=activation, padding="same", kernel_initializer="he_normal")(conv3)
if activation is None:
conv3 = KAF(256, D=kaf_D, kernel=kaf_kernel, init_fcn=init_fcn, conv=True)(conv3)
pool3 = MaxPooling2D(pool_size=(2, 2))(conv3)
conv4 = Conv2D(512, 3, activation=activation, padding="same", kernel_initializer="he_normal")(pool3)
if activation is None:
conv4 = KAF(512, D=kaf_D, kernel=kaf_kernel, init_fcn=init_fcn, conv=True)(conv4)
conv4 = Conv2D(512, 3, activation=activation, padding="same", kernel_initializer="he_normal")(conv4)
if activation is None:
conv4 = KAF(512, D=kaf_D, kernel=kaf_kernel, init_fcn=init_fcn, conv=True)(conv4)
drop4 = Dropout(0.5)(conv4)
pool4 = MaxPooling2D(pool_size=(2, 2))(drop4)
conv5 = Conv2D(1024, 3, activation=activation, padding="same", kernel_initializer="he_normal")(pool4)
if activation is None:
conv5 = KAF(1024, D=kaf_D, kernel=kaf_kernel, init_fcn=init_fcn, conv=True)(conv5)
conv5 = Conv2D(1024, 3, activation=activation, padding="same", kernel_initializer="he_normal")(conv5)
if activation is None:
conv5 = KAF(1024, D=kaf_D, kernel=kaf_kernel, init_fcn=init_fcn, conv=True)(conv5)
drop5 = Dropout(0.5)(conv5)
up6 = Conv2D(512, 2, activation=activation, padding="same", kernel_initializer="he_normal")(
UpSampling2D(size=(2, 2))(drop5)
)
if activation is None:
up6 = KAF(512, D=kaf_D, kernel=kaf_kernel, init_fcn=init_fcn, conv=True)(up6)
merge6 = concatenate([drop4, up6], axis=3)
conv6 = Conv2D(512, 3, activation=activation, padding="same", kernel_initializer="he_normal")(merge6)
if activation is None:
conv6 = KAF(512, D=kaf_D, kernel=kaf_kernel, init_fcn=init_fcn, conv=True)(conv6)
conv6 = Conv2D(512, 3, activation=activation, padding="same", kernel_initializer="he_normal")(conv6)
if activation is None:
conv6 = KAF(512, D=kaf_D, kernel=kaf_kernel, init_fcn=init_fcn, conv=True)(conv6)
up7 = Conv2D(256, 2, activation=activation, padding="same", kernel_initializer="he_normal")(
UpSampling2D(size=(2, 2))(conv6)
)
if activation is None:
up7 = KAF(256, D=kaf_D, kernel=kaf_kernel, init_fcn=init_fcn, conv=True)(up7)
merge7 = concatenate([conv3, up7], axis=3)
conv7 = Conv2D(256, 3, activation=activation, padding="same", kernel_initializer="he_normal")(merge7)
if activation is None:
conv7 = KAF(256, D=kaf_D, kernel=kaf_kernel, init_fcn=init_fcn, conv=True)(conv7)
conv7 = Conv2D(256, 3, activation=activation, padding="same", kernel_initializer="he_normal")(conv7)
if activation is None:
conv7 = KAF(256, D=kaf_D, kernel=kaf_kernel, init_fcn=init_fcn, conv=True)(conv7)
up8 = Conv2D(128, 2, activation=activation, padding="same", kernel_initializer="he_normal")(
UpSampling2D(size=(2, 2))(conv7)
)
if activation is None:
up8 = KAF(128, D=kaf_D, kernel=kaf_kernel, init_fcn=init_fcn, conv=True)(up8)
merge8 = concatenate([conv2, up8], axis=3)
conv8 = Conv2D(128, 3, activation=activation, padding="same", kernel_initializer="he_normal")(merge8)
if activation is None:
conv8 = KAF(128, D=kaf_D, kernel=kaf_kernel, init_fcn=init_fcn, conv=True)(conv8)
conv8 = Conv2D(128, 3, activation=activation, padding="same", kernel_initializer="he_normal")(conv8)
if activation is None:
conv8 = KAF(128, D=kaf_D, kernel=kaf_kernel, init_fcn=init_fcn, conv=True)(conv8)
up9 = Conv2D(64, 2, activation=activation, padding="same", kernel_initializer="he_normal")(
UpSampling2D(size=(2, 2))(conv8)
)
if activation is None:
up9 = KAF(64, D=kaf_D, kernel=kaf_kernel, init_fcn=init_fcn, conv=True)(up9)
merge9 = concatenate([conv1, up9], axis=3)
conv9 = Conv2D(64, 3, activation=activation, padding="same", kernel_initializer="he_normal")(merge9)
if activation is None:
conv9 = KAF(64, D=kaf_D, kernel=kaf_kernel, init_fcn=init_fcn, conv=True)(conv9)
conv9 = Conv2D(64, 3, activation=activation, padding="same", kernel_initializer="he_normal")(conv9)
if activation is None:
conv9 = KAF(64, D=kaf_D, kernel=kaf_kernel, init_fcn=init_fcn, conv=True)(conv9)
conv9 = Conv2D(3, 3, activation=activation, padding="same", kernel_initializer="he_normal")(conv9)
if activation is None:
conv9 = KAF(3, D=kaf_D, kernel=kaf_kernel, init_fcn=init_fcn, conv=True)(conv9)
conv10 = Conv2D(3, 1, activation="softmax")(conv9)
model = Model(input=inputs, output=conv10)
model.summary()
if pretrained_weights:
model.load_weights(pretrained_weights)
return model