-
Notifications
You must be signed in to change notification settings - Fork 11
/
model_e_resunet.py
95 lines (71 loc) · 3.5 KB
/
model_e_resunet.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
# Sept. 2019
# Tran Le Anh, MSc Student
# Satellite Image Processing Lab, Myongji Univ., Yongin, South Korea
# https://sites.google.com/view/leanhtran
# Expanded Residual Unet Model (Architecture)
import numpy as np
import os
import skimage.io as io
import skimage.transform as trans
import numpy as np
from keras.models import *
from keras.layers import *
from keras.optimizers import *
from keras.callbacks import ModelCheckpoint, LearningRateScheduler
from keras import backend as keras
def res_block(in_put, n_kernels):
conv = BatchNormalization(axis=3)(in_put)
conv = Conv2D(n_kernels, 3, activation = 'relu', padding = 'same', kernel_initializer = 'he_normal')(conv)
conv = BatchNormalization(axis=3)(conv)
conv = Conv2D(n_kernels, 3, activation = 'relu', padding = 'same', kernel_initializer = 'he_normal')(conv)
conv_shortcut = Conv2D(n_kernels, 3, activation = 'relu', padding = 'same', kernel_initializer = 'he_normal')(in_put)
conv_shortcut = BatchNormalization(axis=3)(conv_shortcut)
conv = Add()([conv, conv_shortcut])
return conv
def unet(pretrained_weights = None,input_size = (512,512,1)):
inputs = Input(input_size)
# Level 1
conv1 = Conv2D(64, 3, activation = 'relu', padding = 'same', kernel_initializer = 'he_normal')(inputs)
conv1 = BatchNormalization(axis=3)(conv1)
conv1 = Conv2D(64, 3, activation = 'relu', padding = 'same', kernel_initializer = 'he_normal')(conv1)
conv1 = BatchNormalization(axis=3)(conv1)
conv1 = Add()([conv1, inputs])
pool1 = MaxPooling2D(pool_size=(2, 2))(conv1)
# Level 2
conv2 = res_block(pool1, n_kernels = 128)
pool2 = MaxPooling2D(pool_size=(2, 2))(conv2)
# Level 3
conv3 = res_block(pool2, n_kernels = 256)
pool3 = MaxPooling2D(pool_size=(2, 2))(conv3)
# Level 4
conv4 = res_block(pool3, n_kernels = 512)
drop4 = Dropout(0.5)(conv4)
pool4 = MaxPooling2D(pool_size=(2, 2))(drop4)
# Level 5
conv5 = res_block(pool4, n_kernels = 1024)
drop5 = Dropout(0.5)(conv5)
# Level 6
up6 = Conv2D(512, 2, activation = 'relu', padding = 'same', kernel_initializer = 'he_normal')(UpSampling2D(size = (2,2))(drop5))
merge6 = concatenate([drop4,up6], axis = 3)
conv6 = res_block(merge6, n_kernels = 512)
# Level 7
up7 = Conv2D(256, 2, activation = 'relu', padding = 'same', kernel_initializer = 'he_normal')(UpSampling2D(size = (2,2))(conv6))
merge7 = concatenate([conv3, up7], axis = 3)
conv7 = res_block(merge7, n_kernels = 256)
# Level 8
up8 = Conv2D(128, 2, activation = 'relu', padding = 'same', kernel_initializer = 'he_normal')(UpSampling2D(size = (2,2))(conv7))
merge8 = concatenate([conv2, up8], axis = 3)
conv8 = res_block(merge8, n_kernels = 128)
# Level 9
up9 = Conv2D(64, 2, activation = 'relu', padding = 'same', kernel_initializer = 'he_normal')(UpSampling2D(size = (2,2))(conv8))
merge9 = concatenate([conv1, up9], axis = 3)
conv9 = res_block(merge9, n_kernels = 64)
conv9 = Conv2D(2, 3, activation = 'relu', padding = 'same', kernel_initializer = 'he_normal')(conv9)
conv10 = Conv2D(1, 1, activation = 'sigmoid')(conv9)
model = Model(input = inputs, output = conv10)
model.compile(optimizer = Adam(lr = 1e-4), loss = 'binary_crossentropy', metrics = ['accuracy'])
model.summary()
if(pretrained_weights):
model.load_weights(pretrained_weights)
return model