forked from lzhang124/fetal
-
Notifications
You must be signed in to change notification settings - Fork 0
/
models.py
224 lines (167 loc) · 10.6 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
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
import os
import tensorflow as tf
import util
from datetime import datetime
from keras.models import Model
from keras.optimizers import Adam
from keras import backend as K
from keras import layers
from process import uncrop
def dice_coef(y_true, y_pred):
y_true_f = K.flatten(y_true)
y_pred_f = K.flatten(y_pred)
intersection = K.sum(y_true_f * y_pred_f)
return (2. * intersection) / (K.sum(y_true_f) + K.sum(y_pred_f))
def dice_loss(y_true, y_pred):
return 1 - dice_coef(y_true, y_pred)
def weighted_crossentropy(weight=None, boundary_weight=None, pool=3):
w = (.5, .5) if weight is None else weight
epsilon = K.epsilon()
def loss_fn(y_true, y_pred):
y_pred = K.clip(y_pred, epsilon, 1 - epsilon)
cross_entropy = K.stack([-(y_true * K.log(y_pred)), -((1 - y_true) * K.log(1 - y_pred))],
axis=-1)
loss = cross_entropy * w
if boundary_weight is not None:
y_true_avg = K.pool3d(y_true, pool_size=(pool,)*3, padding='same', pool_mode='avg')
boundaries = K.cast(y_true_avg >= epsilon, 'float32') \
* K.cast(y_true_avg <= 1 - epsilon, 'float32')
loss += cross_entropy * K.stack([boundaries, boundaries], axis=-1) * boundary_weight
return K.mean(K.sum(loss, axis=-1))
return loss_fn
class BaseModel:
def __init__(self, input_size, name=None, filename=None):
self.input_size = input_size
self.name = name if name else self.__class__.__name__.lower()
self._new_model()
if filename is not None:
self.model.load_weights(filename)
def _new_model(self):
raise NotImplementedError()
def save(self):
self.model.save('models/{}_weights.{}.h5'.format(self.name, datetime.now().strftime('%m.%d.%y-%H:%M:%S')))
def compile(self, weight):
raise NotImplementedError()
def train(self, generator, val_gen, epochs):
self.model.fit_generator(generator,
epochs=epochs,
validation_data=val_gen,
verbose=1)
def predict(self, generator, path):
preds = self.model.predict_generator(generator, verbose=1)
# FIXME
for i in range(preds.shape[0]):
fname = generator.inputs[i].split('/')[-1]
header = util.header(generator.inputs[i])
util.save_vol(uncrop(preds[i], generator.shape), os.path.join(path, fname), header)
def test(self, generator):
return self.model.evaluate_generator(generator)
class UNet(BaseModel):
def _new_model(self):
inputs = layers.Input(shape=self.input_size)
conv1 = layers.Conv3D(32, (3, 3, 3), activation='relu', padding='same')(inputs)
conv1 = layers.Conv3D(32, (3, 3, 3), activation='relu', padding='same')(conv1)
pool1 = layers.MaxPooling3D(pool_size=(2, 2, 2))(conv1)
conv2 = layers.Conv3D(64, (3, 3, 3), activation='relu', padding='same')(pool1)
conv2 = layers.Conv3D(64, (3, 3, 3), activation='relu', padding='same')(conv2)
pool2 = layers.MaxPooling3D(pool_size=(2, 2, 2))(conv2)
conv3 = layers.Conv3D(128, (3, 3, 3), activation='relu', padding='same')(pool2)
conv3 = layers.Conv3D(128, (3, 3, 3), activation='relu', padding='same')(conv3)
pool3 = layers.MaxPooling3D(pool_size=(2, 2, 2))(conv3)
conv4 = layers.Conv3D(256, (3, 3, 3), activation='relu', padding='same')(pool3)
conv4 = layers.Conv3D(256, (3, 3, 3), activation='relu', padding='same')(conv4)
pool4 = layers.MaxPooling3D(pool_size=(2, 2, 2))(conv4)
conv5 = layers.Conv3D(512, (3, 3, 3), activation='relu', padding='same')(pool4)
conv5 = layers.Conv3D(512, (3, 3, 3), activation='relu', padding='same')(conv5)
up6 = layers.Conv3DTranspose(256, (2, 2, 2), strides=(2, 2, 2), padding='same')(conv5)
conc6 = layers.concatenate([up6, conv4])
conv6 = layers.Conv3D(256, (3, 3, 3), activation='relu', padding='same')(conc6)
conv6 = layers.Conv3D(256, (3, 3, 3), activation='relu', padding='same')(conv6)
up7 = layers.Conv3DTranspose(128, (2, 2, 2), strides=(2, 2, 2), padding='same')(conv6)
conc7 = layers.concatenate([up7, conv3])
conv7 = layers.Conv3D(128, (3, 3, 3), activation='relu', padding='same')(conc7)
conv7 = layers.Conv3D(128, (3, 3, 3), activation='relu', padding='same')(conv7)
up8 = layers.Conv3DTranspose(64, (2, 2, 2), strides=(2, 2, 2), padding='same')(conv7)
conc8 = layers.concatenate([up8, conv2])
conv8 = layers.Conv3D(64, (3, 3, 3), activation='relu', padding='same')(conc8)
conv8 = layers.Conv3D(64, (3, 3, 3), activation='relu', padding='same')(conv8)
up9 = layers.Conv3DTranspose(32, (2, 2, 2), strides=(2, 2, 2), padding='same')(conv8)
conc9 = layers.concatenate([up9, conv1])
conv9 = layers.Conv3D(32, (3, 3, 3), activation='relu', padding='same')(conc9)
conv9 = layers.Conv3D(32, (3, 3, 3), activation='relu', padding='same')(conv9)
outputs = layers.Conv3D(1, (1, 1, 1), activation='sigmoid')(conv9)
self.model = Model(inputs=inputs, outputs=outputs)
def compile(self, weight):
self.model.compile(optimizer=Adam(lr=1e-4),
loss=weighted_crossentropy(weight=weight, boundary_weight=0.2),
metrics=['accuracy', dice_coef])
class UNetSmall(UNet):
def _new_model(self):
inputs = layers.Input(shape=self.input_size)
conv1 = layers.Conv3D(16, (3, 3, 3), activation='relu', padding='same')(inputs)
conv1 = layers.Conv3D(16, (3, 3, 3), activation='relu', padding='same')(conv1)
pool1 = layers.MaxPooling3D(pool_size=(2, 2, 2))(conv1)
conv2 = layers.Conv3D(32, (3, 3, 3), activation='relu', padding='same')(pool1)
conv2 = layers.Conv3D(32, (3, 3, 3), activation='relu', padding='same')(conv2)
pool2 = layers.MaxPooling3D(pool_size=(2, 2, 2))(conv2)
conv3 = layers.Conv3D(64, (3, 3, 3), activation='relu', padding='same')(pool2)
conv3 = layers.Conv3D(64, (3, 3, 3), activation='relu', padding='same')(conv3)
pool3 = layers.MaxPooling3D(pool_size=(2, 2, 2))(conv3)
conv4 = layers.Conv3D(128, (3, 3, 3), activation='relu', padding='same')(pool3)
conv4 = layers.Conv3D(128, (3, 3, 3), activation='relu', padding='same')(conv4)
up5 = layers.Conv3DTranspose(64, (2, 2, 2), strides=(2, 2, 2), padding='same')(conv4)
conc5 = layers.concatenate([up5, conv3])
conv5 = layers.Conv3D(64, (3, 3, 3), activation='relu', padding='same')(conc6)
conv5 = layers.Conv3D(64, (3, 3, 3), activation='relu', padding='same')(conv5)
up6 = layers.Conv3DTranspose(32, (2, 2, 2), strides=(2, 2, 2), padding='same')(conv5)
conc6 = layers.concatenate([up6, conv2])
conv6 = layers.Conv3D(32, (3, 3, 3), activation='relu', padding='same')(conc6)
conv6 = layers.Conv3D(32, (3, 3, 3), activation='relu', padding='same')(conv6)
up7 = layers.Conv3DTranspose(16, (2, 2, 2), strides=(2, 2, 2), padding='same')(conv6)
conc7 = layers.concatenate([up7, conv1])
conv7 = layers.Conv3D(16, (3, 3, 3), activation='relu', padding='same')(conc7)
conv7 = layers.Conv3D(16, (3, 3, 3), activation='relu', padding='same')(conv7)
outputs = layers.Conv3D(1, (1, 1, 1), activation='sigmoid')(conv7)
self.model = Model(inputs=inputs, outputs=outputs)
class UNetBig(UNet):
def _new_model(self):
inputs = layers.Input(shape=self.input_size)
conv1 = layers.Conv3D(16, (3, 3, 3), activation='relu', padding='same')(inputs)
conv1 = layers.Conv3D(16, (3, 3, 3), activation='relu', padding='same')(conv1)
pool1 = layers.MaxPooling3D(pool_size=(2, 2, 2))(conv1)
conv2 = layers.Conv3D(32, (3, 3, 3), activation='relu', padding='same')(pool1)
conv2 = layers.Conv3D(32, (3, 3, 3), activation='relu', padding='same')(conv2)
pool2 = layers.MaxPooling3D(pool_size=(2, 2, 2))(conv2)
conv3 = layers.Conv3D(64, (3, 3, 3), activation='relu', padding='same')(pool2)
conv3 = layers.Conv3D(64, (3, 3, 3), activation='relu', padding='same')(conv3)
pool3 = layers.MaxPooling3D(pool_size=(2, 2, 2))(conv3)
conv4 = layers.Conv3D(128, (3, 3, 3), activation='relu', padding='same')(pool3)
conv4 = layers.Conv3D(128, (3, 3, 3), activation='relu', padding='same')(conv4)
pool4 = layers.MaxPooling3D(pool_size=(2, 2, 2))(conv4)
conv5 = layers.Conv3D(256, (3, 3, 3), activation='relu', padding='same')(pool4)
conv5 = layers.Conv3D(256, (3, 3, 3), activation='relu', padding='same')(conv5)
pool5 = layers.MaxPooling3D(pool_size=(2, 2, 2))(conv5)
conv6 = layers.Conv3D(512, (3, 3, 3), activation='relu', padding='same')(pool5)
conv6 = layers.Conv3D(512, (3, 3, 3), activation='relu', padding='same')(conv6)
up7 = layers.Conv3DTranspose(256, (2, 2, 2), strides=(2, 2, 2), padding='same')(conv6)
conc7 = layers.concatenate([up7, conv5])
conv7 = layers.Conv3D(256, (3, 3, 3), activation='relu', padding='same')(conc7)
conv7 = layers.Conv3D(256, (3, 3, 3), activation='relu', padding='same')(conv7)
up8 = layers.Conv3DTranspose(128, (2, 2, 2), strides=(2, 2, 2), padding='same')(conv7)
conc8 = layers.concatenate([up8, conv4])
conv8 = layers.Conv3D(128, (3, 3, 3), activation='relu', padding='same')(conc8)
conv8 = layers.Conv3D(128, (3, 3, 3), activation='relu', padding='same')(conv8)
up9 = layers.Conv3DTranspose(64, (2, 2, 2), strides=(2, 2, 2), padding='same')(conv8)
conc9 = layers.concatenate([up9, conv3])
conv9 = layers.Conv3D(64, (3, 3, 3), activation='relu', padding='same')(conc9)
conv9 = layers.Conv3D(64, (3, 3, 3), activation='relu', padding='same')(conv9)
up10 = layers.Conv3DTranspose(32, (2, 2, 2), strides=(2, 2, 2), padding='same')(conv9)
conc10 = layers.concatenate([up10, conv2])
conv10 = layers.Conv3D(32, (3, 3, 3), activation='relu', padding='same')(conc10)
conv10 = layers.Conv3D(32, (3, 3, 3), activation='relu', padding='same')(conv10)
up11 = layers.Conv3DTranspose(16, (2, 2, 2), strides=(2, 2, 2), padding='same')(conv10)
conc11 = layers.concatenate([up11, conv1])
conv11 = layers.Conv3D(16, (3, 3, 3), activation='relu', padding='same')(conc11)
conv11 = layers.Conv3D(16, (3, 3, 3), activation='relu', padding='same')(conv11)
outputs = layers.Conv3D(1, (1, 1, 1), activation='sigmoid')(conv11)
self.model = Model(inputs=inputs, outputs=outputs)