-
Notifications
You must be signed in to change notification settings - Fork 0
Expand file tree
/
Copy pathCNN-UNet-Modeli.py
More file actions
202 lines (167 loc) · 9.31 KB
/
CNN-UNet-Modeli.py
File metadata and controls
202 lines (167 loc) · 9.31 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
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
import os
import sys
import random
import numpy as np
import scipy.io as sio
import matplotlib.pyplot as plt
import matplotlib.image as mpimg
from PIL import Image
from sklearn import preprocessing
from sklearn.preprocessing import scale
import tensorflow as tf
from tensorflow import keras
N_train = [2500,2500,2500,2500,2500,2500]
N_test_start = [2500,2500,2500,2500,2500,2500]
N_test = [250,250,250,250,250,250]
N_class = len(N_train)
# Veri boyutlarımızın nasıl olması gerektiğini tanımlıyoruz.
data_sizeX = 200
data_sizeY = 600
mask_sizeX = 256
mask_sizeY = 256
num_channel = 1
# Verileri içerisine aktaracağımız dizi türündeki değişkenleri tanımlıyoruz.
train_data1 = []
train_data2 = []
train_mask= []
test_data1 = []
test_data2 = []
test_mask = []
path = '' # Buraya verilerinizin bilgisayardaki dosya konumu gelmeli
sub_path = ''
# Verileri bilgisayardan çalışmamıza yükleme işlemini gerçekleştiriyoruz.
for i in range(N_class):
for j in range(N_train[i]):
data1 = sio.loadmat(path+'dataset/%d/data1/%d.mat'%(i+1, j+1))['data1']
train_data1.append(data1.reshape((data_sizeX,data_sizeY,1)))
data2 = sio.loadmat(path+'dataset/%d/data2/%d.mat'%(i+1, j+1))['data2']
train_data2.append(data2.reshape((data_sizeX,data_sizeY,1)))
mask = sio.loadmat(path+'dataset/%d/mask/%d.mat'%(i+1, j+1))['mask']
train_mask.append(mask.reshape((mask_sizeX,mask_sizeY,1)))
for j in range(N_test_start[i], N_test_start[i]+N_test[i]):
data1 = sio.loadmat(path+'dataset/%d/data1/%d.mat'%(i+1, j+1))['data1']
test_data1.append(data1.reshape((data_sizeX,data_sizeY,1)))
data2 = sio.loadmat(path+'dataset/%d/data2/%d.mat'%(i+1, j+1))['data2']
test_data2.append(data2.reshape((data_sizeX,data_sizeY,1)))
mask = sio.loadmat(path+'dataset/%d/mask/%d.mat'%(i+1, j+1))['mask']
test_mask.append(mask.reshape((mask_sizeX,mask_sizeY,1)))
# Verileri dizi türündeki değişkenlere aktarıyoruz. Numpy dizileri verilerde daha kolaylıkla işlem yapmamıza olanak sağlamaktadır.
train_data1 = np.array(train_data1)
train_data2 = np.array(train_data2)
train_mask = np.array(train_mask)
test_data1 = np.array(test_data1)
test_data2 = np.array(test_data2)
test_mask = np.array(test_mask)
print("veriler basariyla okundu...")
def create_unet_model():
f0 = 16
f = [f0, f0*2, f0*4, f0*8, f0*16, f0*32]
inputs1 = keras.layers.Input((data_sizeX, data_sizeY, num_channel))
inputs2 = keras.layers.Input((data_sizeX, data_sizeY, num_channel))
p10 = inputs1
p11 = keras.layers.Conv2D(f[0], (3, 3), padding='same', activation="relu", strides=1)(p10)
p11 = keras.layers.Conv2D(f[0], (3,3), padding='same', strides=1,activation="relu")(p11)
p11=keras.layers.MaxPool2D((2, 2), (2, 2))(p11)
p12 = keras.layers.Conv2D(f[1], (3, 3), padding='same', activation="relu", strides=1)(p11)
p12 = keras.layers.Conv2D(f[1], (3,3), padding='same', strides=1,activation="relu")(p12)
p12=keras.layers.MaxPool2D((2, 2), (2, 2))(p12)
p13 = keras.layers.Conv2D(f[2], (3, 3), padding='same', activation="relu", strides=1)(p12)
p13 = keras.layers.Conv2D(f[2], (3,3), padding='same', strides=1)(p13)
p13=keras.layers.MaxPool2D((2, 2), (2, 2))(p13)
p14 = keras.layers.Conv2D(f[3], (3, 3), padding='same', activation="relu", strides=1)(p13)
p14 = keras.layers.Conv2D(f[3], (3,3), padding='same', strides=1)(p14)
p14=keras.layers.MaxPool2D((2, 2), (2, 2))(p14)
p15 =keras.layers.Conv2D(f[4],(3,3),padding="same",activation="relu",strides=1)(p14)
p15 =keras.layers.Conv2D(f[4],(3,3),padding="same",activation="relu",strides=1)(p15)
p15= keras.layers.MaxPool2D((2, 2), (2, 2))(p15)
p20 = inputs2
p21 = keras.layers.Conv2D(f[0], (3, 3), padding='same', activation="relu", strides=1)(p20)
p21 = keras.layers.Conv2D(f[0], (3,3), padding='same', strides=1,activation="relu")(p21)
p21 = keras.layers.MaxPool2D((2, 2), (2, 2))(p21)
p22 = keras.layers.Conv2D(f[1], (3, 3), padding='same', activation="relu", strides=1)(p21)
p22 = keras.layers.Conv2D(f[1], (3,3), padding='same', strides=1,activation="relu")(p22)
p22 = keras.layers.MaxPool2D((2, 2), (2, 2))(p22)
p23 = keras.layers.Conv2D(f[2], (3, 3), padding='same', activation="relu", strides=1)(p22)
p23 = keras.layers.Conv2D(f[2], (3,3), padding='same', strides=1)(p23)
p23=keras.layers.MaxPool2D((2, 2), (2, 2))(p23)
p24 = keras.layers.Conv2D(f[3], (3, 3), padding='same', activation="relu", strides=1)(p23)
p24 = keras.layers.Conv2D(f[3], (3,3), padding='same', strides=1)(p24)
p24=keras.layers.MaxPool2D((2, 2), (2, 2))(p24)
p25 =keras.layers.Conv2D(f[4],(3,3),padding="same",activation="relu",strides=1)(p24)
p25 =keras.layers.Conv2D(f[4],(3,3),padding="same",activation="relu",strides=1)(p25)
p25= keras.layers.MaxPool2D((2, 2), (2, 2))(p25)
# Fusion
v1 = keras.layers.Conv2D(f[4], (1, 1), padding='same', strides=1)(p15)
q2 = keras.layers.Conv2D(f[4], (1, 1), padding='same', strides=1)(p25)
a1 = keras.layers.MultiHeadAttention(num_heads=8, key_dim=f[4]//8)(q2, v1)
o1 = keras.layers.LayerNormalization()(a1 + v1)
fu1 = keras.layers.LayerNormalization()(o1 + keras.layers.Conv2D(f[4], (1, 1), padding='same', strides=1)(o1))
v2 = keras.layers.Conv2D(f[4], (1, 1), padding='same', strides=1)(p25)
q1 = keras.layers.Conv2D(f[4], (1, 1), padding='same', strides=1)(p15)
a2 = keras.layers.MultiHeadAttention(num_heads=8, key_dim=f[4]//8)(q1, v2)
o2 = keras.layers.LayerNormalization()(a2 + v2)
fu2 = keras.layers.LayerNormalization()(o2 + keras.layers.Conv2D(f[4], (1, 1), padding='same', strides=1)(o2))
#concatenate
c = keras.layers.Conv2D(f[5], (3,3), padding='same', strides=(1,3))(keras.layers.Concatenate()([fu1,fu2]))
c = keras.layers.Activation('relu')(c)
c = keras.layers.Conv2DTranspose(f[5], (3,3), padding='valid', strides=1)(c)
c = keras.layers.Activation('relu')(c)
c = keras.layers.Conv2D(f[5], (3,3), padding='same', strides=1)(c)
fu = keras.layers.Activation('relu')(c)
# Decoder
us1 = keras.layers.UpSampling2D((2, 2))(fu)
c1 = keras.layers.Conv2D(f[4], (3, 3), padding='same', activation="relu",strides=1)(us1)
c1 = keras.layers.Conv2D(f[4], (3, 3), padding='same', activation="relu",strides=1)(c1)
us2 = keras.layers.UpSampling2D((2, 2))(c1)
c2 = keras.layers.Conv2D(f[3], (3, 3), padding='same', activation="relu",strides=1)(us2)
c2 = keras.layers.Conv2D(f[3], (3, 3), padding='same', activation="relu",strides=1)(c2)
us3 = keras.layers.UpSampling2D((2, 2))(c2)
c3 = keras.layers.Conv2D(f[2], (3, 3), padding='same', activation="relu",strides=1)(us3)
c3 = keras.layers.Conv2D(f[2], (3, 3), padding='same', activation="relu",strides=1)(c3)
us4 = keras.layers.UpSampling2D((2, 2))(c3)
c4 = keras.layers.Conv2D(f[1], (3, 3), padding='same', activation= "relu",strides=1)(us4)
c4 = keras.layers.Conv2D(f[1], (3, 3), padding='same', activation="relu",strides=1)(c4)
us5 = keras.layers.UpSampling2D((2, 2))(c4)
c5 = keras.layers.Conv2D(f[0], (3, 3), padding='same', activation= "relu",strides=1)(us5)
c5 = keras.layers.Conv2D(f[0], (3, 3), padding='same', activation="relu",strides=1)(c5)
outputs = keras.layers.Conv2D(1, (1, 1), padding='same')(c5)
model = tf.keras.Model([inputs1, inputs2], outputs)
return model
"""
def connect_block(x, filters, kernel_size=(3,3)):
c = keras.layers.Conv2D(filters, kernel_size, padding='same', strides=(1,3))(x)
c = keras.layers.Activation('relu')(c)
c = keras.layers.Conv2DTranspose(filters, kernel_size, padding='valid', strides=1)(c)
c = keras.layers.Activation('relu')(c)
c = keras.layers.Conv2D(filters, kernel_size, padding='same', strides=1)(c)
c = keras.layers.Activation('relu')(c)
return c
def metric(optimizer):
def learning_rate(y_true, y_pred):
return optimizer.learning_rate
return learning_rate
"""
model = create_unet_model()
model.summary()
print("model basariyla olusturuldu")
#train
"""
total_epoch = 1
batch_size = 10
model_path = path + sub_path + 'model.h5'
Adam = keras.optimizers.Adam(learning_rate=1e-4)
lr_metric = metric(Adam)
model.compile(optimizer=Adam, loss='mse', metrics=[lr_metric])
model_checkpoint = keras.callbacks.ModelCheckpoint(model_path, monitor='val_loss', verbose=1, save_best_only=True)
lr_checkpoint = keras.callbacks.ReduceLROnPlateau(monitor='val_loss', factor=0.98, patience=1, min_lr=0)
history = model.fit(x=[train_data1, train_data2], y=train_mask, batch_size=batch_size, epochs=total_epoch, verbose=2, \
validation_data=([test_data1, test_data2], test_mask), callbacks=[model_checkpoint, lr_checkpoint])
# Testing
model.load_weights(model_path)
model.evaluate(x=[test_data1,test_data2], y=test_mask, batch_size=batch_size)
test_pred = model.predict([test_data1,test_data2])
sio.savemat(path+sub_path+'data1.mat', {'data1': test_data1})
sio.savemat(path+sub_path+'data2.mat', {'data2': test_data2})
sio.savemat(path+sub_path+'mask.mat', {'mask': test_mask})
sio.savemat(path+sub_path+'pred.mat', {'pred': test_pred})
print("model ciktilari basariyla kaydedildi")"""