-
Notifications
You must be signed in to change notification settings - Fork 0
/
model.py
159 lines (143 loc) · 7.59 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
#DL-B Model
import numpy as np
import keras
from keras import backend as K
from keras.utils.data_utils import get_file
from keras.models import Sequential, Model
from keras.layers.core import Flatten, Dense, Dropout, Lambda,Reshape
from keras.layers import Input,LeakyReLU, Dense, Layer,LocallyConnected1D,Activation
from keras.layers.convolutional import Convolution2D, MaxPooling2D, ZeroPadding2D,AveragePooling2D
from keras.layers import Conv1D,Conv2D, Conv3D, Conv2DTranspose,ConvLSTM2D, SimpleRNN, LSTM, Permute
from keras.layers import UpSampling2D, merge,Reshape
from keras.optimizers import SGD, RMSprop, Adam
from keras.preprocessing import image
from keras.layers.normalization import BatchNormalization
from keras import optimizers
from keras import regularizers
from keras.engine import InputLayer
from keras.callbacks import ModelCheckpoint
#Initialise kernel for nearest neighbour interpolation
def kernel_init(shape):
"""Initialise kernel for nearest neighbour interpolation to be used in the deep learning model architecture
shape : convolution layer filter size.
"""
kernel = np.ones(shape)
return kernel
def get_dlB(inputShape, chN=3, lr = 1e-4, loss='binary_crossentropy', metrics=['accuracy'], modelFile=""):
"""Getting dlB model architecture with predefined input shape
inputShape : tupple of input shape [x,y].
modelFile : path to the pretrained model file [string]. If left empty, no model will be loaded.
Returns : keras model with loaded pretrained model.
"""
inputs = Input((inputShape[0],inputShape[1],chN))
inputsNormed=BatchNormalization()(inputs)
conv1a = Conv2D(16, 3, padding = 'same', kernel_initializer = 'he_normal')(inputsNormed)
conv1a = BatchNormalization()(conv1a)
conv1a = Activation('relu')(conv1a)
conv1b = Conv2D(16, 3, padding = 'same', kernel_initializer = 'he_normal')(conv1a)
conv1b = BatchNormalization()(conv1b)
conv1b = Activation('relu')(conv1b)
conv1Res1 = keras.layers.Concatenate()([conv1a,conv1b])
pool1 = AveragePooling2D(pool_size=(2, 2))(conv1Res1)
conv2a = Conv2D(32, 3, padding = 'same', kernel_initializer = 'he_normal')(pool1)
conv2a = BatchNormalization()(conv2a)
conv2a = Activation('relu')(conv2a)
conv2b = Conv2D(32, 3, padding = 'same', kernel_initializer = 'he_normal')(conv2a)
conv2b = BatchNormalization()(conv2b)
conv2b = Activation('relu')(conv2b)
conv2Res1 = keras.layers.Concatenate()([conv2a,conv2b])
pool2 = AveragePooling2D(pool_size=(2, 2))(conv2Res1)
conv3a = Conv2D(64, 3, padding = 'same', kernel_initializer = 'he_normal')(pool2)
conv3a = BatchNormalization()(conv3a)
conv3a = Activation('relu')(conv3a)
conv3b = Conv2D(64, 3, padding = 'same', kernel_initializer = 'he_normal')(conv3a)
conv3b = BatchNormalization()(conv3b)
conv3b = Activation('relu')(conv3b)
conv3Res1 = keras.layers.Concatenate()([conv3a,conv3b])
pool3 = AveragePooling2D(pool_size=(2, 2))(conv3Res1)
conv4a = Conv2D(128, 3, padding = 'same', kernel_initializer = 'he_normal')(pool3)
conv4a = BatchNormalization()(conv4a)
conv4a = Activation('relu')(conv4a)
conv4b = Conv2D(128, 3, padding = 'same', kernel_initializer = 'he_normal')(conv4a)
conv4b = BatchNormalization()(conv4b)
conv4b = Activation('relu')(conv4b)
conv4Res1 = keras.layers.Concatenate()([conv4a,conv4b])
drop4 = Dropout(0.5)(conv4Res1)
pool4 = AveragePooling2D(pool_size=(2, 2))(drop4)
conv5a = Conv2D(256, 3, padding = 'same', kernel_initializer = 'he_normal')(pool4)
conv5a = BatchNormalization()(conv5a)
conv5a = Activation('relu')(conv5a)
conv5b = Conv2D(256, 3, padding = 'same', kernel_initializer = 'he_normal')(conv5a)
conv5b = BatchNormalization()(conv5b)
conv5b = Activation('relu')(conv5b)
conv5Res1 = keras.layers.Concatenate()([conv5a,conv5b])
drop5 = Dropout(0.5)(conv5Res1)
fcConv1a = Conv2D(8, 1, padding = 'same', kernel_initializer = 'he_normal')(drop5)
fcConv1a = BatchNormalization()(fcConv1a)
fcConv1a = Activation('relu')(fcConv1a)
dropFcConv1a = Dropout(0.5)(fcConv1a)
fc1a = Flatten()(dropFcConv1a)
fc1b = Dense(256, activation='relu')(fc1a)
dropFc1 = Dropout(0.5)(fc1b)
fc1c = Dense(1, activation='sigmoid')(dropFc1)
up6 = Conv2D(128, 2, padding = 'same', kernel_initializer = 'he_normal')(UpSampling2D(size = (2,2))(drop5))
up6 = BatchNormalization()(up6)
up6 = Activation('relu')(up6)
merge6 = keras.layers.Concatenate()([conv4Res1,up6])
conv6a = Conv2D(128, 3, padding = 'same', kernel_initializer = 'he_normal')(merge6)
conv6a = BatchNormalization()(conv6a)
conv6a = Activation('relu')(conv6a)
conv6b = Conv2D(128, 3, padding = 'same', kernel_initializer = 'he_normal')(conv6a)
conv6b = BatchNormalization()(conv6b)
conv6b = Activation('relu')(conv6b)
conv6Res1 = keras.layers.Concatenate()([conv6a,conv6b])
up7 = Conv2D(64, 2, padding = 'same', kernel_initializer = 'he_normal')(UpSampling2D(size = (2,2))(conv6Res1))
up7 = BatchNormalization()(up7)
up7 = Activation('relu')(up7)
merge7 = keras.layers.Concatenate()([conv3Res1,up7])
conv7a = Conv2D(64, 3, padding = 'same', kernel_initializer = 'he_normal')(merge7)
conv7a = BatchNormalization()(conv7a)
conv7a = Activation('relu')(conv7a)
conv7b = Conv2D(64, 3, padding = 'same', kernel_initializer = 'he_normal')(conv7a)
conv7b = BatchNormalization()(conv7b)
conv7b = Activation('relu')(conv7b)
conv7Res1 = keras.layers.Concatenate()([conv7a,conv7b])
up8 = Conv2D(32, 2, padding = 'same', kernel_initializer = 'he_normal')(UpSampling2D(size = (2,2))(conv7Res1))
up8 = BatchNormalization()(up8)
up8 = Activation('relu')(up8)
merge8 = keras.layers.Concatenate()([conv2Res1,up8])
conv8a = Conv2D(32, 3, padding = 'same', kernel_initializer = 'he_normal')(merge8)
conv8a = BatchNormalization()(conv8a)
conv8a = Activation('relu')(conv8a)
conv8b = Conv2D(32, 3, padding = 'same', kernel_initializer = 'he_normal')(conv8a)
conv8b = BatchNormalization()(conv8b)
conv8b = Activation('relu')(conv8b)
conv8Res1 = keras.layers.Concatenate()([conv8a,conv8b])
up9 = Conv2D(16, 2, padding = 'same', kernel_initializer = 'he_normal')(UpSampling2D(size = (2,2))(conv8Res1))
up9 = BatchNormalization()(up9)
up9 = Activation('relu')(up9)
merge9 = keras.layers.Concatenate()([conv1Res1,up9])
conv9a = Conv2D(16, 3, padding = 'same', kernel_initializer = 'he_normal')(merge9)
conv9a = BatchNormalization()(conv9a)
conv9a = Activation('relu')(conv9a)
conv9b = Conv2D(16, 3, padding = 'same', kernel_initializer = 'he_normal')(conv9a)
conv9b = BatchNormalization()(conv9b)
conv9b = Activation('relu')(conv9b)
conv9Res1 = keras.layers.Concatenate()([conv9a,conv9b])
conv9 = Conv2D(2, 3, padding = 'same', kernel_initializer = 'he_normal')(conv9Res1)
conv9 = BatchNormalization()(conv9)
conv9 = Activation('relu')(conv9)
conv10 = Conv2D(1, 1)(conv9)
conv10 = BatchNormalization()(conv10)
conv10 = Activation('sigmoid')(conv10)
fc1cFR = Conv2DTranspose(1,inputShape,padding=('valid'),kernel_initializer=kernel_init,trainable=False)(Reshape((1,1,1))(fc1c))
finalLayer = keras.layers.Multiply()([conv10,fc1cFR])
model = Model(input = inputs, output = [finalLayer,fc1c])
model.compile(optimizer = Adam(lr = 1e-4), loss = loss, metrics = metrics)
if modelFile!="":
try:
model.load_weights(modelFile) #Model B
print("Model loaded: "+str(modelFile))
except NameError:
print("Can't find model, no model loaded")
return model