-
Notifications
You must be signed in to change notification settings - Fork 3
/
model.py
350 lines (288 loc) · 23.5 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
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
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
# This U-Net implementation is originally imported from zhixuhao's 'unet' GitHub repository and modified for
# 3D convolutions instead of 3D convolutions.
# https://github.com/zhixuhao/unet/blob/master/model.py
from tensorflow.keras.models import Model
from tensorflow.keras.layers import (
Input,
Conv3D,
MaxPooling3D,
UpSampling3D,
Dropout,
Conv3DTranspose,
BatchNormalization,
concatenate
)
def unet_3d_upsampling_dropout(input_size=(240, 240, 160, 4), unet_resize_factor=2, unet_dropout_rate=0.3, num_classes=4,
binary_model=False):
"""Constructs a U-Net 3D segmentation model with Dropout layers and UpSampling3D -> Conv3D layers.
Args:
input_size: (tuple) Keras model input shape is (batch_size, height, width, length, channels) with
'channels_last', (default: (240, 240, 160, 4)). Note: depth must be a multiple of 16.
Source: 'data_format' parameter documentation: https://keras.io/api/layers/convolution_layers/convolution3d/
unet_resize_factor: (int) Resize factor of the number of filters (channels) per Convolutional layer in the U-Net
model (must be >= 1, such that 1 means retaining the original number of filters (channels)
per Convolutional layer in the U-Net model) (default: 2 (half-size))
unet_dropout_rate: (float) Dropout rate for the Dropout layers in the U-Net model (default: 0.3).
num_classes: (int) Number of classes in the training dataset (default: 4).
binary_model: (boolean) If True, make the last layer have one filter with 'sigmoid' activation for a 3D binary
segmentation model.
"""
inputs = Input(shape=input_size)
# Contractive path
conv1 = Conv3D(filters=64 // unet_resize_factor, kernel_size=3, activation='relu', padding='same', kernel_initializer='he_normal')(inputs)
conv1 = Conv3D(filters=64 // unet_resize_factor, kernel_size=3, activation='relu', padding='same', kernel_initializer='he_normal')(conv1)
pool1 = MaxPooling3D(pool_size=(2, 2, 2))(conv1)
conv2 = Conv3D(filters=128 // unet_resize_factor, kernel_size=3, activation='relu', padding='same', kernel_initializer='he_normal')(pool1)
conv2 = Conv3D(filters=128 // unet_resize_factor, kernel_size=3, activation='relu', padding='same', kernel_initializer='he_normal')(conv2)
pool2 = MaxPooling3D(pool_size=(2, 2, 2))(conv2)
conv3 = Conv3D(filters=256 // unet_resize_factor, kernel_size=3, activation='relu', padding='same', kernel_initializer='he_normal')(pool2)
conv3 = Conv3D(filters=256 // unet_resize_factor, kernel_size=3, activation='relu', padding='same', kernel_initializer='he_normal')(conv3)
pool3 = MaxPooling3D(pool_size=(2, 2, 2))(conv3)
conv4 = Conv3D(filters=512 // unet_resize_factor, kernel_size=3, activation='relu', padding='same', kernel_initializer='he_normal')(pool3)
conv4 = Conv3D(filters=512 // unet_resize_factor, kernel_size=3, activation='relu', padding='same', kernel_initializer='he_normal')(conv4)
drop4 = Dropout(rate=unet_dropout_rate)(conv4)
pool4 = MaxPooling3D(pool_size=(2, 2, 2))(drop4)
conv5 = Conv3D(filters=1024 // unet_resize_factor, kernel_size=3, activation='relu', padding='same', kernel_initializer='he_normal')(pool4)
conv5 = Conv3D(filters=1024 // unet_resize_factor, kernel_size=3, activation='relu', padding='same', kernel_initializer='he_normal')(conv5)
drop5 = Dropout(rate=unet_dropout_rate)(conv5)
# Expansive path
up6 = Conv3D(filters=512 // unet_resize_factor, kernel_size=2, activation='relu', padding='same', kernel_initializer='he_normal')(UpSampling3D(size=(2, 2, 2))(drop5))
merge6 = concatenate([drop4, up6], axis=4)
conv6 = Conv3D(filters=512 // unet_resize_factor, kernel_size=3, activation='relu', padding='same', kernel_initializer='he_normal')(merge6)
conv6 = Conv3D(filters=512 // unet_resize_factor, kernel_size=3, activation='relu', padding='same', kernel_initializer='he_normal')(conv6)
up7 = Conv3D(filters=256 // unet_resize_factor, kernel_size=2, activation='relu', padding='same', kernel_initializer='he_normal')(UpSampling3D(size=(2, 2, 2))(conv6))
merge7 = concatenate([conv3, up7], axis=4)
conv7 = Conv3D(filters=256 // unet_resize_factor, kernel_size=3, activation='relu', padding='same', kernel_initializer='he_normal')(merge7)
conv7 = Conv3D(filters=256 // unet_resize_factor, kernel_size=3, activation='relu', padding='same', kernel_initializer='he_normal')(conv7)
up8 = Conv3D(filters=128 // unet_resize_factor, kernel_size=2, activation='relu', padding='same', kernel_initializer='he_normal')(UpSampling3D(size=(2, 2, 2))(conv7))
merge8 = concatenate([conv2, up8], axis=4)
conv8 = Conv3D(filters=128 // unet_resize_factor, kernel_size=3, activation='relu', padding='same', kernel_initializer='he_normal')(merge8)
conv8 = Conv3D(filters=128 // unet_resize_factor, kernel_size=3, activation='relu', padding='same', kernel_initializer='he_normal')(conv8)
up9 = Conv3D(filters=64 // unet_resize_factor, kernel_size=2, activation='relu', padding='same', kernel_initializer='he_normal')(UpSampling3D(size=(2, 2, 2))(conv8))
merge9 = concatenate([conv1, up9], axis=4)
conv9 = Conv3D(filters=64 // unet_resize_factor, kernel_size=3, activation='relu', padding='same', kernel_initializer='he_normal')(merge9)
conv9 = Conv3D(filters=64 // unet_resize_factor, kernel_size=3, activation='relu', padding='same', kernel_initializer='he_normal')(conv9)
conv9 = Conv3D(filters=2, kernel_size=3, activation='relu', padding='same', kernel_initializer='he_normal')(conv9)
# Final layer
if binary_model:
conv10 = Conv3D(filters=1, kernel_size=1, activation="sigmoid")(conv9)
else:
conv10 = Conv3D(filters=num_classes, kernel_size=1, activation="softmax")(conv9)
model = Model(inputs=inputs, outputs=conv10)
return model
def unet_3d_conv3dtranspose_dropout(input_size=(240, 240, 160, 4), unet_resize_factor=2, unet_dropout_rate=0.3, num_classes=4,
binary_model=False):
"""Constructs a U-Net 3D segmentation model with Dropout layers and Conv3DTranspose layers instead of
UpSampling3D -> Conv3D layers.
Args:
input_size: (tuple) Keras model input shape is (batch_size, height, width, length, channels) with
'channels_last', (default: (240, 240, 160, 4)). Note: depth must be a multiple of 16.
Source: 'data_format' parameter documentation: https://keras.io/api/layers/convolution_layers/convolution3d/
unet_resize_factor: (int) Resize factor of the number of filters (channels) per Convolutional layer in the U-Net
model (must be >= 1, such that 1 means retaining the original number of filters (channels)
per Convolutional layer in the U-Net model) (default: 2 (half-size))
unet_dropout_rate: (float) Dropout rate for the Dropout layers in the U-Net model (default: 0.3).
num_classes: (int) Number of classes in the training dataset (default: 4).
binary_model: (boolean) If True, make the last layer have one filter with 'sigmoid' activation for a 3D binary
segmentation model.
"""
inputs = Input(shape=input_size)
# Contractive path
conv1 = Conv3D(filters=64 // unet_resize_factor, kernel_size=3, activation='relu', padding='same', kernel_initializer='he_normal')(inputs)
conv1 = Conv3D(filters=64 // unet_resize_factor, kernel_size=3, activation='relu', padding='same', kernel_initializer='he_normal')(conv1)
pool1 = MaxPooling3D(pool_size=(2, 2, 2))(conv1)
conv2 = Conv3D(filters=128 // unet_resize_factor, kernel_size=3, activation='relu', padding='same', kernel_initializer='he_normal')(pool1)
conv2 = Conv3D(filters=128 // unet_resize_factor, kernel_size=3, activation='relu', padding='same', kernel_initializer='he_normal')(conv2)
pool2 = MaxPooling3D(pool_size=(2, 2, 2))(conv2)
conv3 = Conv3D(filters=256 // unet_resize_factor, kernel_size=3, activation='relu', padding='same', kernel_initializer='he_normal')(pool2)
conv3 = Conv3D(filters=256 // unet_resize_factor, kernel_size=3, activation='relu', padding='same', kernel_initializer='he_normal')(conv3)
pool3 = MaxPooling3D(pool_size=(2, 2, 2))(conv3)
conv4 = Conv3D(filters=512 // unet_resize_factor, kernel_size=3, activation='relu', padding='same', kernel_initializer='he_normal')(pool3)
conv4 = Conv3D(filters=512 // unet_resize_factor, kernel_size=3, activation='relu', padding='same', kernel_initializer='he_normal')(conv4)
drop4 = Dropout(rate=unet_dropout_rate)(conv4)
pool4 = MaxPooling3D(pool_size=(2, 2, 2))(drop4)
conv5 = Conv3D(filters=1024 // unet_resize_factor, kernel_size=3, activation='relu', padding='same', kernel_initializer='he_normal')(pool4)
conv5 = Conv3D(filters=1024 // unet_resize_factor, kernel_size=3, activation='relu', padding='same', kernel_initializer='he_normal')(conv5)
drop5 = Dropout(rate=unet_dropout_rate)(conv5)
# Expansive path
up6 = Conv3DTranspose(filters=512 // unet_resize_factor, kernel_size=(2, 2, 2), strides=(2, 2, 2), padding="same", kernel_initializer='he_normal')(drop5)
merge6 = concatenate([drop4, up6], axis=4)
conv6 = Conv3D(filters=512 // unet_resize_factor, kernel_size=3, activation='relu', padding='same', kernel_initializer='he_normal')(merge6)
conv6 = Conv3D(filters=512 // unet_resize_factor, kernel_size=3, activation='relu', padding='same', kernel_initializer='he_normal')(conv6)
up7 = Conv3DTranspose(filters=128 // unet_resize_factor, kernel_size=(2, 2, 2), strides=(2, 2, 2), padding="same", kernel_initializer='he_normal')(conv6)
merge7 = concatenate([conv3, up7], axis=4)
conv7 = Conv3D(filters=256 // unet_resize_factor, kernel_size=3, activation='relu', padding='same', kernel_initializer='he_normal')(merge7)
conv7 = Conv3D(filters=256 // unet_resize_factor, kernel_size=3, activation='relu', padding='same', kernel_initializer='he_normal')(conv7)
up8 = Conv3DTranspose(filters=64 // unet_resize_factor, kernel_size=(2, 2, 2), strides=(2, 2, 2), padding="same", kernel_initializer='he_normal')(conv7)
merge8 = concatenate([conv2, up8], axis=4)
conv8 = Conv3D(filters=128 // unet_resize_factor, kernel_size=3, activation='relu', padding='same', kernel_initializer='he_normal')(merge8)
conv8 = Conv3D(filters=128 // unet_resize_factor, kernel_size=3, activation='relu', padding='same', kernel_initializer='he_normal')(conv8)
up9 = Conv3DTranspose(filters=32 // unet_resize_factor, kernel_size=(2, 2, 2), strides=(2, 2, 2), padding="same", kernel_initializer='he_normal')(conv8)
merge9 = concatenate([conv1, up9], axis=4)
conv9 = Conv3D(filters=64 // unet_resize_factor, kernel_size=3, activation='relu', padding='same', kernel_initializer='he_normal')(merge9)
conv9 = Conv3D(filters=64 // unet_resize_factor, kernel_size=3, activation='relu', padding='same', kernel_initializer='he_normal')(conv9)
conv9 = Conv3D(filters=2, kernel_size=3, activation='relu', padding='same', kernel_initializer='he_normal')(conv9)
# Final layer
if binary_model:
conv10 = Conv3D(filters=1, kernel_size=1, activation="sigmoid")(conv9)
else:
conv10 = Conv3D(filters=num_classes, kernel_size=1, activation="softmax")(conv9)
model = Model(inputs=inputs, outputs=conv10)
return model
def unet_3d_upsampling_batchnormalization(input_size=(240, 240, 160, 4), unet_resize_factor=2, num_classes=4, binary_model=False):
"""Constructs a U-Net 3D segmentation model with BatchNormalization layers after each Conv3D layer instead of
using Dropout layers in the expansive path and with using UpSampling3D -> Conv3D layers.
Args:
input_size: (tuple) Keras model input shape is (batch_size, height, width, length, channels) with
'channels_last', (default: (240, 240, 160, 4)). Note: depth must be a multiple of 16.
Source: 'data_format' parameter documentation: https://keras.io/api/layers/convolution_layers/convolution3d/
unet_resize_factor: (int) Resize factor of the number of filters (channels) per Convolutional layer in the U-Net
model (must be >= 1, such that 1 means retaining the original number of filters (channels)
per Convolutional layer in the U-Net model) (default: 2 (half-size)).
num_classes: (int) Number of classes in the training dataset (default: 4).
binary_model: (boolean) If True, make the last layer have one filter with 'sigmoid' activation for a 3D binary
segmentation model.
"""
inputs = Input(shape=input_size)
# Contractive path
conv1 = Conv3D(filters=64 // unet_resize_factor, kernel_size=3, activation='relu', padding='same', kernel_initializer='he_normal')(inputs)
bn1 = BatchNormalization()(conv1)
conv1 = Conv3D(filters=64 // unet_resize_factor, kernel_size=3, activation='relu', padding='same', kernel_initializer='he_normal')(bn1)
bn1 = BatchNormalization()(conv1)
pool1 = MaxPooling3D(pool_size=(2, 2, 2))(bn1)
conv2 = Conv3D(filters=128 // unet_resize_factor, kernel_size=3, activation='relu', padding='same', kernel_initializer='he_normal')(pool1)
bn2 = BatchNormalization()(conv2)
conv2 = Conv3D(filters=128 // unet_resize_factor, kernel_size=3, activation='relu', padding='same', kernel_initializer='he_normal')(bn2)
bn2 = BatchNormalization()(conv2)
pool2 = MaxPooling3D(pool_size=(2, 2, 2))(bn2)
conv3 = Conv3D(filters=256 // unet_resize_factor, kernel_size=3, activation='relu', padding='same', kernel_initializer='he_normal')(pool2)
bn3 = BatchNormalization()(conv3)
conv3 = Conv3D(filters=256 // unet_resize_factor, kernel_size=3, activation='relu', padding='same', kernel_initializer='he_normal')(bn3)
bn3 = BatchNormalization()(conv3)
pool3 = MaxPooling3D(pool_size=(2, 2, 2))(bn3)
conv4 = Conv3D(filters=512 // unet_resize_factor, kernel_size=3, activation='relu', padding='same', kernel_initializer='he_normal')(pool3)
bn4 = BatchNormalization()(conv4)
conv4 = Conv3D(filters=512 // unet_resize_factor, kernel_size=3, activation='relu', padding='same', kernel_initializer='he_normal')(bn4)
bn4 = BatchNormalization()(conv4)
pool4 = MaxPooling3D(pool_size=(2, 2, 2))(bn4)
conv5 = Conv3D(filters=1024 // unet_resize_factor, kernel_size=3, activation='relu', padding='same', kernel_initializer='he_normal')(pool4)
bn5 = BatchNormalization()(conv5)
conv5 = Conv3D(filters=1024 // unet_resize_factor, kernel_size=3, activation='relu', padding='same', kernel_initializer='he_normal')(bn5)
bn5 = BatchNormalization()(conv5)
# Expansive path
up6 = Conv3D(filters=512 // unet_resize_factor, kernel_size=2, activation='relu', padding='same', kernel_initializer='he_normal')(UpSampling3D(size=(2, 2, 2))(bn5))
bn6 = BatchNormalization()(up6)
merge6 = concatenate([bn4, bn6], axis=4)
conv6 = Conv3D(filters=512 // unet_resize_factor, kernel_size=3, activation='relu', padding='same', kernel_initializer='he_normal')(merge6)
bn6 = BatchNormalization()(conv6)
conv6 = Conv3D(filters=512 // unet_resize_factor, kernel_size=3, activation='relu', padding='same', kernel_initializer='he_normal')(bn6)
bn6 = BatchNormalization()(conv6)
up7 = Conv3D(filters=256 // unet_resize_factor, kernel_size=2, activation='relu', padding='same', kernel_initializer='he_normal')(UpSampling3D(size=(2, 2, 2))(bn6))
bn7 = BatchNormalization()(up7)
merge7 = concatenate([conv3, bn7], axis=4)
conv7 = Conv3D(filters=256 // unet_resize_factor, kernel_size=3, activation='relu', padding='same', kernel_initializer='he_normal')(merge7)
bn7 = BatchNormalization()(conv7)
conv7 = Conv3D(filters=256 // unet_resize_factor, kernel_size=3, activation='relu', padding='same', kernel_initializer='he_normal')(bn7)
bn7 = BatchNormalization()(conv7)
up8 = Conv3D(filters=128 // unet_resize_factor, kernel_size=2, activation='relu', padding='same', kernel_initializer='he_normal')(UpSampling3D(size=(2, 2, 2))(bn7))
bn8 = BatchNormalization()(up8)
merge8 = concatenate([conv2, bn8], axis=4)
conv8 = Conv3D(filters=128 // unet_resize_factor, kernel_size=3, activation='relu', padding='same', kernel_initializer='he_normal')(merge8)
bn8 = BatchNormalization()(conv8)
conv8 = Conv3D(filters=128 // unet_resize_factor, kernel_size=3, activation='relu', padding='same', kernel_initializer='he_normal')(bn8)
bn8 = BatchNormalization()(conv8)
up9 = Conv3D(filters=64 // unet_resize_factor, kernel_size=2, activation='relu', padding='same', kernel_initializer='he_normal')(UpSampling3D(size=(2, 2, 2))(bn8))
bn9 = BatchNormalization()(up9)
merge9 = concatenate([conv1, bn9], axis=4)
conv9 = Conv3D(filters=64 // unet_resize_factor, kernel_size=3, activation='relu', padding='same', kernel_initializer='he_normal')(merge9)
bn9 = BatchNormalization()(conv9)
conv9 = Conv3D(filters=64 // unet_resize_factor, kernel_size=3, activation='relu', padding='same', kernel_initializer='he_normal')(bn9)
bn9 = BatchNormalization()(conv9)
conv9 = Conv3D(filters=2, kernel_size=3, activation='relu', padding='same', kernel_initializer='he_normal')(bn9)
bn9 = BatchNormalization()(conv9)
# Final layer
if binary_model:
conv10 = Conv3D(filters=1, kernel_size=1, activation="sigmoid")(bn9)
else:
conv10 = Conv3D(filters=num_classes, kernel_size=1, activation="softmax")(bn9)
model = Model(inputs=inputs, outputs=conv10)
return model
def unet_3d_conv3dtranspose_batchnormalization(input_size=(240, 240, 160, 4), unet_resize_factor=2, num_classes=4, binary_model=False):
"""Constructs a U-Net 3D segmentation model with BatchNormalization layers after each Conv3D layer instead of
using Dropout layers in the expansive path and with using Conv3DTranspose layers instead of UpSampling3D -> Conv3D
layers.
Args:
input_size: (tuple) Keras model input shape is (batch_size, height, width, length, channels) with
'channels_last', (default: (240, 240, 160, 4)). Note: depth must be a multiple of 16.
Source: 'data_format' parameter documentation: https://keras.io/api/layers/convolution_layers/convolution3d/
unet_resize_factor: (int) Resize factor of the number of filters (channels) per Convolutional layer in the U-Net
model (must be >= 1, such that 1 means retaining the original number of filters (channels)
per Convolutional layer in the U-Net model) (default: 2 (half-size)).
num_classes: (int) Number of classes in the training dataset (default: 4).
binary_model: (boolean) If True, make the last layer have one filter with 'sigmoid' activation for a 3D binary
segmentation model.
"""
inputs = Input(shape=input_size)
# Contractive path
conv1 = Conv3D(filters=64 // unet_resize_factor, kernel_size=3, activation='relu', padding='same', kernel_initializer='he_normal')(inputs)
bn1 = BatchNormalization()(conv1)
conv1 = Conv3D(filters=64 // unet_resize_factor, kernel_size=3, activation='relu', padding='same', kernel_initializer='he_normal')(bn1)
bn1 = BatchNormalization()(conv1)
pool1 = MaxPooling3D(pool_size=(2, 2, 2))(bn1)
conv2 = Conv3D(filters=128 // unet_resize_factor, kernel_size=3, activation='relu', padding='same', kernel_initializer='he_normal')(pool1)
bn2 = BatchNormalization()(conv2)
conv2 = Conv3D(filters=128 // unet_resize_factor, kernel_size=3, activation='relu', padding='same', kernel_initializer='he_normal')(bn2)
bn2 = BatchNormalization()(conv2)
pool2 = MaxPooling3D(pool_size=(2, 2, 2))(bn2)
conv3 = Conv3D(filters=256 // unet_resize_factor, kernel_size=3, activation='relu', padding='same', kernel_initializer='he_normal')(pool2)
bn3 = BatchNormalization()(conv3)
conv3 = Conv3D(filters=256 // unet_resize_factor, kernel_size=3, activation='relu', padding='same', kernel_initializer='he_normal')(bn3)
bn3 = BatchNormalization()(conv3)
pool3 = MaxPooling3D(pool_size=(2, 2, 2))(bn3)
conv4 = Conv3D(filters=512 // unet_resize_factor, kernel_size=3, activation='relu', padding='same', kernel_initializer='he_normal')(pool3)
bn4 = BatchNormalization()(conv4)
conv4 = Conv3D(filters=512 // unet_resize_factor, kernel_size=3, activation='relu', padding='same', kernel_initializer='he_normal')(bn4)
bn4 = BatchNormalization()(conv4)
pool4 = MaxPooling3D(pool_size=(2, 2, 2))(bn4)
conv5 = Conv3D(filters=1024 // unet_resize_factor, kernel_size=3, activation='relu', padding='same', kernel_initializer='he_normal')(pool4)
bn5 = BatchNormalization()(conv5)
conv5 = Conv3D(filters=1024 // unet_resize_factor, kernel_size=3, activation='relu', padding='same', kernel_initializer='he_normal')(bn5)
bn5 = BatchNormalization()(conv5)
# Expansive path
up6 = Conv3DTranspose(filters=512 // unet_resize_factor, kernel_size=(2, 2, 2), strides=(2, 2, 2), padding="same", kernel_initializer='he_normal')(bn5)
bn6 = BatchNormalization()(up6)
merge6 = concatenate([bn4, bn6], axis=4)
conv6 = Conv3D(filters=512 // unet_resize_factor, kernel_size=3, activation='relu', padding='same', kernel_initializer='he_normal')(merge6)
bn6 = BatchNormalization()(conv6)
conv6 = Conv3D(filters=512 // unet_resize_factor, kernel_size=3, activation='relu', padding='same', kernel_initializer='he_normal')(bn6)
bn6 = BatchNormalization()(conv6)
up7 = Conv3DTranspose(filters=128 // unet_resize_factor, kernel_size=(2, 2, 2), strides=(2, 2, 2), padding="same", kernel_initializer='he_normal')(bn6)
bn7 = BatchNormalization()(up7)
merge7 = concatenate([conv3, bn7], axis=4)
conv7 = Conv3D(filters=256 // unet_resize_factor, kernel_size=3, activation='relu', padding='same', kernel_initializer='he_normal')(merge7)
bn7 = BatchNormalization()(conv7)
conv7 = Conv3D(filters=256 // unet_resize_factor, kernel_size=3, activation='relu', padding='same', kernel_initializer='he_normal')(bn7)
bn7 = BatchNormalization()(conv7)
up8 = Conv3DTranspose(filters=64 // unet_resize_factor, kernel_size=(2, 2, 2), strides=(2, 2, 2), padding="same", kernel_initializer='he_normal')(bn7)
bn8 = BatchNormalization()(up8)
merge8 = concatenate([conv2, bn8], axis=4)
conv8 = Conv3D(filters=128 // unet_resize_factor, kernel_size=3, activation='relu', padding='same', kernel_initializer='he_normal')(merge8)
bn8 = BatchNormalization()(conv8)
conv8 = Conv3D(filters=128 // unet_resize_factor, kernel_size=3, activation='relu', padding='same', kernel_initializer='he_normal')(bn8)
bn8 = BatchNormalization()(conv8)
up9 = Conv3DTranspose(filters=32 // unet_resize_factor, kernel_size=(2, 2, 2), strides=(2, 2, 2), padding="same", kernel_initializer='he_normal')(bn8)
bn9 = BatchNormalization()(up9)
merge9 = concatenate([conv1, bn9], axis=4)
conv9 = Conv3D(filters=64 // unet_resize_factor, kernel_size=3, activation='relu', padding='same', kernel_initializer='he_normal')(merge9)
bn9 = BatchNormalization()(conv9)
conv9 = Conv3D(filters=64 // unet_resize_factor, kernel_size=3, activation='relu', padding='same', kernel_initializer='he_normal')(bn9)
bn9 = BatchNormalization()(conv9)
conv9 = Conv3D(filters=2, kernel_size=3, activation='relu', padding='same', kernel_initializer='he_normal')(bn9)
bn9 = BatchNormalization()(conv9)
# Final layer
if binary_model:
conv10 = Conv3D(filters=1, kernel_size=1, activation="sigmoid")(bn9)
else:
conv10 = Conv3D(filters=num_classes, kernel_size=1, activation="softmax")(bn9)
model = Model(inputs=inputs, outputs=conv10)
return model