-
Notifications
You must be signed in to change notification settings - Fork 3
Expand file tree
/
Copy pathmodel2D.py
More file actions
128 lines (109 loc) · 5.72 KB
/
model2D.py
File metadata and controls
128 lines (109 loc) · 5.72 KB
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
from keras.models import *
from keras.layers import *
from keras.optimizers import *
import numpy as np
from keras.losses import *
from DL_utils import metrics
def unet2D(pretrained_weights,input_size, num_channels, loss_function, base_n_filters = 64, learning_rate=0.001):
print("started UNet")
inputs = Input(input_size)
conv1 = Conv2D(base_n_filters, 3, padding='same', kernel_initializer='he_normal')(inputs)
conv1 = LeakyReLU(alpha=0.2)(conv1)
conv1 = BatchNormalization()(conv1)
print("finished 1. convolution")
conv1 = Conv2D(base_n_filters, 3, padding='same', kernel_initializer='he_normal')(conv1)
conv1 = LeakyReLU(alpha=0.2)(conv1)
conv1 = BatchNormalization()(conv1)
drop1 = Dropout(0.1)(conv1)
pool1 = MaxPooling2D(pool_size=(2, 2))(drop1)
conv2 = Conv2D(base_n_filters*2, 3, padding='same', kernel_initializer='he_normal')(pool1)
conv2 = LeakyReLU(alpha=0.2)(conv2)
conv2 = BatchNormalization()(conv2)
conv2 = Conv2D(base_n_filters*2, 3, padding='same', kernel_initializer='he_normal')(conv2)
conv2 = LeakyReLU(alpha=0.2)(conv2)
conv2 = BatchNormalization()(conv2)
drop2 = Dropout(0.1)(conv2)
pool2 = MaxPooling2D(pool_size=(2, 2))(drop2)
conv3 = Conv2D(base_n_filters*4, 3, padding='same', kernel_initializer='he_normal')(pool2)
conv3 = LeakyReLU(alpha=0.2)(conv3)
conv3 = BatchNormalization()(conv3)
conv3 = Conv2D(base_n_filters*4, 3, padding='same', kernel_initializer='he_normal')(conv3)
conv3 = LeakyReLU(alpha=0.2)(conv3)
conv3 = BatchNormalization()(conv3)
drop3 = Dropout(0.1)(conv3)
pool3 = MaxPooling2D(pool_size=(2, 2))(drop3)
conv4 = Conv2D(base_n_filters*8, 3, padding='same', kernel_initializer='he_normal')(pool3)
conv4 = LeakyReLU(alpha=0.2)(conv4)
conv4 = BatchNormalization()(conv4)
conv4 = Conv2D(base_n_filters*8, 3, padding='same', kernel_initializer='he_normal')(conv4)
conv4 = LeakyReLU(alpha=0.2)(conv4)
conv4 = BatchNormalization()(conv4)
drop4 = Dropout(0.1)(conv4)
pool4 = MaxPooling2D(pool_size=(2, 2))(drop4)
conv5 = Conv2D(base_n_filters*16, 3, padding='same', kernel_initializer='he_normal')(pool4)
conv5 = LeakyReLU(alpha=0.2)(conv5)
conv5 = BatchNormalization()(conv5)
conv5 = Conv2D(base_n_filters*16, 3, padding='same', kernel_initializer='he_normal')(conv5)
conv5 = LeakyReLU(alpha=0.2)(conv5)
conv5 = BatchNormalization()(conv5)
drop5 = Dropout(0.1)(conv5)
up6 = Conv2D(base_n_filters*8, 2, padding='same', kernel_initializer='he_normal')(UpSampling2D(size=(2, 2))(drop5))
merge6 = concatenate([drop4, up6], axis=3)
conv6 = Conv2D(base_n_filters*8, 3,activation = 'relu', padding='same', kernel_initializer='he_normal')(merge6)
conv6 = LeakyReLU(alpha=0.1)(conv6)
conv6 = BatchNormalization()(conv6)
conv6 = Conv2D(base_n_filters*8, 3, activation = 'relu', padding='same', kernel_initializer='he_normal')(conv6)
conv6 = LeakyReLU(alpha=0.1)(conv6)
conv6 = BatchNormalization()(conv6)
up7 = Conv2D(base_n_filters*4, 2, padding='same', kernel_initializer='he_normal')(UpSampling2D(size=(2, 2))(conv6))
merge7 = concatenate([drop3, up7], axis=3)
conv7 = Conv2D(base_n_filters*4, 3, padding='same', kernel_initializer='he_normal')(merge7)
conv7 = LeakyReLU(alpha=0.1)(conv7)
conv7 = BatchNormalization()(conv7)
conv7 = Conv2D(base_n_filters*4, 3, padding='same', kernel_initializer='he_normal')(conv7)
conv7 = LeakyReLU(alpha=0.1)(conv7)
conv7 = BatchNormalization()(conv7)
up8 = Conv2D(base_n_filters*2, 2, padding='same', kernel_initializer='he_normal')(UpSampling2D(size=(2, 2))(conv7))
merge8 = concatenate([drop2, up8], axis=3)
conv8 = Conv2D(base_n_filters*2, 3, padding='same', kernel_initializer='he_normal')(merge8)
conv8 = LeakyReLU(alpha=0.1)(conv8)
conv8 = BatchNormalization()(conv8)
conv8 = Conv2D(base_n_filters*2, 3, padding='same', kernel_initializer='he_normal')(conv8)
conv8 = LeakyReLU(alpha=0.1)(conv8)
conv8 = BatchNormalization()(conv8)
up9 = Conv2D(base_n_filters, 2, padding='same', kernel_initializer='he_normal')(UpSampling2D(size=(2, 2))(conv8))
merge9 = concatenate([drop1, up9], axis=3)
conv9 = Conv2D(base_n_filters, 3, padding='same', kernel_initializer='he_normal')(merge9)
conv9 = LeakyReLU(alpha=0.1)(conv9)
conv9 = BatchNormalization()(conv9)
conv9 = Conv2D(base_n_filters, 3, padding='same', kernel_initializer='he_normal')(conv9)
conv9 = LeakyReLU(alpha=0.1)(conv9)
conv9 = BatchNormalization()(conv9)
conv9 = Conv2D(3, 3,padding='same', kernel_initializer='he_normal')(conv9)
conv9 = LeakyReLU(alpha=0.1)(conv9)
conv9 = BatchNormalization()(conv9)
conv10 = Conv2D(1, 1, activation='sigmoid')(conv9) #Changed activiation from Relu to linear
if loss_function == "dice_loss":
loss_function = eval(loss_function)
model = Model(inputs, conv10) # model = Model(input=inputs, output=conv10)
print("shape input UNet:", np.shape(inputs))
print("shape output UNet:", np.shape(conv10))
model.compile(optimizer=Adam(lr = learning_rate), loss = loss_function, metrics=['mae'])
#To Do: Try beta_1 value of 0.5 as in ocemoglus paper
'''model.compile(optimizer = Adam(lr = 0.001, beta_1 = 0.9, beta_2 = 0.999, epsilon = 1e-8), loss = 'categorical_crossentropy', metrics=['accuracy'])
Other possible losses =
- binary_crossentropy
- categorical_crossentropy
- sparse_categorical_crossentropy
- mean_squared_error
- dice_coef_loss
- total_variaton_loss_mse
- mean_absolute_error
- mean_squared_logarithmic_error
metrics =
- 'accuracy'
- dice_coef '''
#model.summary()
if pretrained_weights:
model.load_weights(pretrained_weights)
return model