-
Notifications
You must be signed in to change notification settings - Fork 15
/
Copy pathtrain_model_up.py
247 lines (221 loc) · 10.6 KB
/
train_model_up.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
# -*- coding: utf-8 -*-
"""
Created on Tue Feb 4 23:30:22 2020
@author: Leo
"""
import os
os.environ["CUDA_VISIBLE_DEVICES"] = "0"
import tensorflow as tf
from keras.backend.tensorflow_backend import set_session
config = tf.ConfigProto()
config.gpu_options.allow_growth = True
set_session(tf.Session(config=config))
# 主函数
#import os
import cv2
import numpy as np
import keras.backend as K
import matplotlib.pyplot as plt
from sklearn.utils import shuffle
from keras.preprocessing import image
from keras.utils.np_utils import to_categorical
from keras.models import Model, load_model
from keras.optimizers import Adam
from keras.layers.core import Activation,Dense
from keras.layers import GlobalMaxPooling2D, GlobalAveragePooling2D, ZeroPadding2D, Cropping2D
from keras.layers import Conv2D, MaxPooling2D, PReLU, concatenate,Input, UpSampling2D
from keras.layers.normalization import BatchNormalization
from keras.callbacks import ModelCheckpoint, EarlyStopping
def img_dilating(img, kernel_size):
'''图像膨胀'''
if kernel_size !=0:
kernel = np.ones((kernel_size,kernel_size),np.uint8)
img = cv2.dilate(img, kernel)
return img
def array_img_reshape(array, target_shape):
'''array图像缩放，要求三通道'''
array = image.img_to_array(image.array_to_img(array).resize(target_shape))
return array
def batch_generater(data_path, data_list, batch_size, shape, n_label):
offset = 0
while True:
load_list = shuffle(data_list)
X = np.zeros((batch_size, *shape))
Y = np.zeros((batch_size, *shape[0:2], n_label))
Z = np.zeros((batch_size, n_label))
for i in range(batch_size):
img_x_path = data_path[0] + '/' + load_list[i + offset]
img_y_path = data_path[1] + '/' + load_list[i + offset]
img_class = int(load_list[i + offset][0])
# load_img, img_to_array
img_x = image.load_img(img_x_path, target_size = shape[0:2])
img_y = image.load_img(img_y_path, target_size = shape[0:2])
img_x = image.img_to_array(img_x)
img_y = image.img_to_array(img_y)
# 对label裂纹做膨胀和缩放
#img_y = img_dilating(img_y, kernel_size=4) # kernel_size=0 时不膨胀
#img_y = array_img_reshape(img_y, (62,62))
# 取单通道, Normalization
img_x = img_x[...,2:]
img_y = img_y[...,2:]
img_x = 1/255.0 * img_x
img_y = 1/255.0 * img_y
# one-hot 编码
img_class = to_categorical(img_class,n_label)
img_y = to_categorical(img_y, n_label)
X[i,...] = img_x
Y[i,...] = img_y
Z[i,...] = img_class
if i+offset >= len(load_list)-1:
load_list = shuffle(data_list)
offset = 0
yield (X, {'seg_output': Y, 'class_output': Z})
offset += batch_size
def conv(x, n_kernel, s_kernel, name_id, training=True, bn=True):
x = Conv2D(n_kernel, s_kernel, strides=(1,1), padding='same',
kernel_initializer='he_normal', activation='relu',
name=str(name_id), trainable=training)(x)
if bn:
x = BatchNormalization()(x)
return x
def build_model_up(input_shape, n_label):
input_img = Input(shape=input_shape)
# segmentation network
seg_conv1 = conv(input_img, 32, 5, 'seg_conv1_01')
seg_conv1 = conv(seg_conv1, 32, 5, 'seg_conv1_02') # 500
seg_maxp1 = MaxPooling2D(pool_size=(2,2))(seg_conv1) # 250
seg_conv2 = conv(seg_maxp1, 64, 5, 'seg_conv2_01') # 250
seg_conv2 = conv(seg_conv2, 64, 5, 'seg_conv2_02')
seg_conv2 = conv(seg_conv2, 64, 5, 'seg_conv2_03') # 250
seg_maxp2 = MaxPooling2D(pool_size=(2,2))(seg_conv2) # 125
seg_conv3 = conv(seg_maxp2, 64, 5, 'seg_conv3_01') # 125
seg_conv3 = conv(seg_conv3, 64, 5, 'seg_conv3_02')
seg_conv3 = conv(seg_conv3, 64, 5, 'seg_conv3_03')
seg_conv3 = conv(seg_conv3, 64, 5, 'seg_conv3_04') # 125
seg_maxp3 = MaxPooling2D(pool_size=(2,2))(seg_conv3) # 62
seg_conv4 = conv(seg_maxp3, 1024, 5, 'seg_conv4_01') # 62
seg_mask = Conv2D(n_label, 1, strides=(1,1), padding='same',
kernel_initializer='he_normal', activation='softmax',
name='seg_mask')(seg_conv4)
seg_upsp5 = UpSampling2D()(seg_conv4) # 124
seg_upsp5 = ZeroPadding2D(padding=((0, 1), (0, 1)))(seg_upsp5) # 125
seg_upsp5 = concatenate([seg_upsp5,seg_conv3],axis=3)
seg_conv5 = conv(seg_upsp5, 64, 5, 'seg_conv5_01')
#seg_conv5 = conv(seg_upsp5, 64, 5, 'seg_conv5_02')
#seg_conv5 = conv(seg_upsp5, 64, 5, 'seg_conv5_03')
#seg_conv5 = conv(seg_upsp5, 64, 5, 'seg_conv5_04') # 125
seg_upsp6 = UpSampling2D()(seg_conv5) # 256
seg_upsp6 = concatenate([seg_upsp6,seg_conv2],axis=3)
seg_conv6 = conv(seg_upsp6, 64, 5, 'seg_conv6_01')
#seg_conv6 = conv(seg_upsp6, 64, 5, 'seg_conv6_02')
#seg_conv6 = conv(seg_upsp6, 64, 5, 'seg_conv6_03') # 250
seg_upsp7 = UpSampling2D()(seg_conv6) # 512
seg_upsp7 = concatenate([seg_upsp7,seg_conv1],axis=3)
seg_conv7 = conv(seg_upsp7, 32, 5, 'seg_conv7_01')
#seg_conv7 = conv(seg_upsp7, 32, 5, 'seg_conv7_02') # 500
seg_output = Conv2D(n_label, 1, strides=(1,1), padding='same',
kernel_initializer='he_normal', activation='softmax',
name='seg_output')(seg_conv7)
# decision network
dec_conv1 = concatenate([seg_conv4,seg_mask],axis=3)
dec_conv1 = MaxPooling2D(pool_size=(2,2))(dec_conv1)
dec_conv2 = conv(dec_conv1, 16, 5, 'dec_conv2_01')
dec_conv2 = MaxPooling2D(pool_size=(2,2))(dec_conv2)
dec_conv3 = conv(dec_conv2, 32, 5, 'dec_conv3_01')
dec_conv3 = MaxPooling2D(pool_size=(2,2))(dec_conv3)
dec_conv4 = conv(dec_conv3, 64, 5, 'dec_conv4_01')
# classification
class1 = GlobalMaxPooling2D()(dec_conv4)
class2 = GlobalAveragePooling2D()(dec_conv4)
class3 = GlobalAveragePooling2D()(seg_mask)
class4 = GlobalMaxPooling2D()(seg_mask)
classes = concatenate([class1,class2,class3,class4],axis=1)
class_output = Dense(n_label, activation='softmax', name='class_output')(classes)
model = Model(inputs=input_img, outputs=[seg_output,class_output])
return model
def show_result(img_id, input_shape):
src_path = './dataset/val_set/src/1_'+str(img_id)+'.png'
lab_path = './dataset/val_set/label/1_'+str(img_id)+'.png'
img1 = image.load_img(src_path, target_size = input_shape[0:2]) # src原图
img2 = image.load_img(lab_path, target_size = input_shape[0:2]) # label原图
img = image.img_to_array(img1)[...,2:]
img = img/255
img = np.expand_dims(img,axis=0)
result = model.predict(img)
fig = result[0][0,...,1]
fig1 = result[0][0,...]
fig1 = np.reshape(fig1,(1,500*500,2))
fig1 = np.argmax(fig1,axis=2).astype(np.int8)
fig1 = np.reshape(fig1,(500,500))
print('Result:',np.rint(result[1][0]))
plt.figure(figsize=(8,8))
plt.subplot(221)
plt.imshow(img1)
plt.subplot(222)
plt.imshow(img2)
plt.subplot(223)
plt.imshow(fig)
plt.subplot(224)
plt.imshow(fig1)
plt.show()
return None
if __name__ == '__main__':
input_shape = (500,500,1)
n_label = 2
batch_size = 2
epoch = 5
train_set_path = ['./dataset/train_set/src','./dataset/train_set/label']
valid_set_path = ['./dataset/val_set/src' ,'./dataset/val_set/label' ]
model_savepath = './model/'
model_savename = 'best_model_up_500'
train_name_list = os.listdir(train_set_path[0])
valid_name_list = os.listdir(valid_set_path[0])
model = build_model_up(input_shape, n_label)
adam = Adam(lr=0.001, epsilon=1e-08, decay=1e-5, amsgrad=True)
model.compile(optimizer=adam,
metrics= {'seg_output': 'mse', 'class_output': 'acc'},
loss = {'seg_output': 'mse', 'class_output': 'binary_crossentropy'},
loss_weights= {'seg_output': 1.0, 'class_output': 0.5})
'''
训练步骤：
1. activation='softmax', loss = {'seg_output':'mse', 'class_output':'binary_crossentropy'}, epoch = 10
2. activation='sigmoid', loss = {'seg_output':'binary_crossentropy', 'class_output':'binary_crossentropy'}, epoch = 10
'''
model.summary()
# 保存模型网络结构
with open(model_savepath+model_savename+'.json', 'w') as f:
f.write(model.to_json())
# 以加载权重的方式加载之前训练的结果
if os.listdir(model_savepath):
try:
model.load_weights(model_savepath + model_savename + '_weights.h5')
print('Model weights loaded!')
except:
print('Model weights load filed!')
save_best_1 = ModelCheckpoint(model_savepath + model_savename + '_weights.h5', monitor='val_seg_output_loss',
verbose=1, save_best_only=True, save_weights_only=True, mode='auto')
save_best_2 = ModelCheckpoint(model_savepath + model_savename + '_weights.h5', monitor='val_class_output_acc',
verbose=1, save_best_only=True, save_weights_only=True, mode='auto')
early_stop = EarlyStopping(monitor='val_loss', min_delta=0, patience=3, verbose=0, mode='auto')
h = model.fit_generator(generator=batch_generater(train_set_path,train_name_list,batch_size,input_shape,n_label),
steps_per_epoch = len(train_name_list)//batch_size,
epochs=epoch, verbose=1, callbacks=[save_best_1, save_best_2, early_stop],
validation_steps = len(valid_name_list)//8,
validation_data=batch_generater(valid_set_path,valid_name_list,8,input_shape,n_label))
plt.figure(1) # 图 1 画 seg_output_mse
plt.plot(h.history['seg_output_mean_squared_error'])
plt.plot(h.history['val_seg_output_mean_squared_error'])
plt.title('seg_output_mse')
plt.ylabel('loss')
plt.xlabel('epoch')
plt.legend(['train', 'validation'], loc='upper right')
plt.savefig('train_val_loss.jpg',dpi=600)
plt.figure(2) # 图 2 画 class_output_acc
plt.plot(h.history['class_output_acc'])
plt.plot(h.history['val_class_output_acc'])
plt.title('class_output_acc')
plt.ylabel('acc')
plt.xlabel('epoch')
plt.legend(['train', 'validation'], loc='upper right')
plt.savefig('train_val_acc.jpg',dpi=600)
show_result(np.random.randint(0,500),input_shape)