-
Notifications
You must be signed in to change notification settings - Fork 4
/
train.py
306 lines (279 loc) · 9.91 KB
/
train.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
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
import os
import numpy as np
import matplotlib.pyplot as plt
import tensorflow as tf
from tensorflow.keras.layers import Input, Dense, Conv2D, Activation, MaxPool2D
from tensorflow.keras.layers import *
from tensorflow.keras.models import Model, Sequential
from tensorflow.keras.optimizers import Adam, SGD
import sys
import glob
import argparse
from keras import __version__
from keras.applications.xception import Xception, preprocess_input
from keras.models import *
from keras.layers import *
from keras.activations import *
from keras.preprocessing.image import ImageDataGenerator
from keras.optimizers import SGD
from keras import optimizers
from keras import callbacks
import skimage.io as io
import skimage.transform as trans
import numpy as np
from keras.callbacks import ModelCheckpoint, LearningRateScheduler
from keras import backend as keras
from keras.regularizers import l2,l1
import pandas as pd
import keras_preprocessing
from keras_preprocessing import image
from keras_preprocessing.image import ImageDataGenerator
import shutil
input_shape = (224,224,3)
nbr_of_classes=38
TRAINING_DIR = "train"
training_datagen = ImageDataGenerator(rotation_range=40,
width_shift_range=0.1,
height_shift_range=0.1,
preprocessing_function=preprocess_input,
horizontal_flip=False,
fill_mode='nearest')
VALIDATION_DIR = "val"
validation_datagen = ImageDataGenerator(preprocessing_function=preprocess_input)
def train():
train_generator = training_datagen.flow_from_directory(
TRAINING_DIR,
target_size=(224,224),
shuffle = True,
class_mode='categorical',
batch_size = 16
)
while True:
x, y = train_generator.next()
yield x, [y, y]
def valid():
validation_generator = validation_datagen.flow_from_directory(
VALIDATION_DIR,
target_size=(224,224),
class_mode='categorical',
shuffle = True,
batch_size= 16
)
while True:
x, y = validation_generator.next()
yield x, [y, y]
#Encoder Start
base_model1 = tf.keras.applications.Xception(include_top=False, weights='imagenet',input_shape = input_shape)
x1_0 = base_model1.output
x1_0 = Flatten(name='Flatten1')(x1_0)
dense1 = Dense(256, name='fc1',activation='relu')(x1_0)
x = classif_out_encoder1 = Dense(38, name='out1', activation = 'softmax')(dense1) # Latent Representation / Bottleneck
#Get Xception's tensors for skip connection.
conv14 = base_model1.get_layer('block14_sepconv2_act').output
conv13 = base_model1.get_layer('block13_sepconv2_bn').output
conv12 = base_model1.get_layer('block12_sepconv3_bn').output
conv11 = base_model1.get_layer('block11_sepconv3_bn').output
conv10 = base_model1.get_layer('block10_sepconv3_bn').output
conv9 = base_model1.get_layer('block9_sepconv3_bn').output
conv8 = base_model1.get_layer('block8_sepconv3_bn').output
conv7 = base_model1.get_layer('block7_sepconv3_bn').output
conv6 = base_model1.get_layer('block6_sepconv3_bn').output
conv5 = base_model1.get_layer('block5_sepconv3_bn').output
conv4 = base_model1.get_layer('block4_sepconv2_bn').output
conv3 = base_model1.get_layer('block3_sepconv2_bn').output
conv2 = base_model1.get_layer('block2_sepconv2_bn').output
conv1 = base_model1.get_layer('block1_conv2_act').output
#Decoder Start
dense2 = Dense(256, activation='relu')(x)
x = Add(name='first_merge')([dense1, dense2])
x = Dense(7*7*2048)(x)
reshape1 = Reshape((7, 7, 2048))(x)
#BLOCK 1
x = SeparableConv2D(2048, (3,3), padding='same', name='block14_start')(reshape1)
x = BatchNormalization()(x)
x = Activation('relu')(x)
x = concatenate([conv14, x], axis = 3)
x = SeparableConv2D(1536, (3,3), padding='same')(x)
x = BatchNormalization()(x)
x = c14 = Activation('relu')(x)
#BLOCK 2
x = UpSampling2D((2,2))(x)
x = Activation('relu')(x)
x = SeparableConv2D(1024, (3,3), padding='same', name='block13_start')(x)
x = BatchNormalization()(x)
x = concatenate([conv13, x], axis = 3)
x = Activation('relu')(x)
x = SeparableConv2D(728, (3,3), padding='same')(x)
x = BatchNormalization()(x)
c1314 = Conv2D(728, (1,1))(UpSampling2D()(c14))
x = add1 = Add()([c1314, x])
#BLOCK 3
x = Activation('relu')(x)
x = SeparableConv2D(728, (3,3), padding='same', name='blockmiddle_start')(x)
x = BatchNormalization()(x)
x = concatenate([conv12, x], axis = 3)
x = Activation('relu')(x)
x = SeparableConv2D(728, (3,3), padding='same')(x)
x = BatchNormalization()(x)
x = Activation('relu')(x)
x = SeparableConv2D(728, (3,3), padding='same')(x)
x = BatchNormalization()(x)
x = add2 = Add()([add1, x])
#BLOCK 4
x = Activation('relu')(x)
x = SeparableConv2D(728, (3,3), padding='same')(x)
x = BatchNormalization()(x)
x = concatenate([conv11, x], axis = 3)
x = Activation('relu')(x)
x = SeparableConv2D(728, (3,3), padding='same')(x)
x = BatchNormalization()(x)
x = Activation('relu')(x)
x = SeparableConv2D(728, (3,3), padding='same')(x)
x = BatchNormalization()(x)
x = add3 = Add()([add2, x])
#BLOCK 5
x = Activation('relu')(x)
x = SeparableConv2D(728, (3,3), padding='same')(x)
x = BatchNormalization()(x)
x = concatenate([conv10, x], axis = 3)
x = Activation('relu')(x)
x = SeparableConv2D(728, (3,3), padding='same')(x)
x = BatchNormalization()(x)
x = Activation('relu')(x)
x = SeparableConv2D(728, (3,3), padding='same')(x)
x = BatchNormalization()(x)
x = add4 = Add()([add3, x])
#BLOCK 6
x = Activation('relu')(x)
x = SeparableConv2D(728, (3,3), padding='same')(x)
x = BatchNormalization()(x)
x = concatenate([conv9, x], axis = 3)
x = Activation('relu')(x)
x = SeparableConv2D(728, (3,3), padding='same')(x)
x = BatchNormalization()(x)
x = Activation('relu')(x)
x = SeparableConv2D(728, (3,3), padding='same')(x)
x = BatchNormalization()(x)
x = add5 = Add()([add4, x])
#BLOCK 7
x = Activation('relu')(x)
x = SeparableConv2D(728, (3,3), padding='same')(x)
x = BatchNormalization()(x)
x = concatenate([conv8, x], axis = 3)
x = Activation('relu')(x)
x = SeparableConv2D(728, (3,3), padding='same')(x)
x = BatchNormalization()(x)
x = Activation('relu')(x)
x = SeparableConv2D(728, (3,3), padding='same')(x)
x = BatchNormalization()(x)
x = add6 = Add()([add5, x])
#BLOCK 8
x = Activation('relu')(x)
x = SeparableConv2D(728, (3,3), padding='same')(x)
x = BatchNormalization()(x)
x = concatenate([conv7, x], axis = 3)
x = Activation('relu')(x)
x = SeparableConv2D(728, (3,3), padding='same')(x)
x = BatchNormalization()(x)
x = Activation('relu')(x)
x = SeparableConv2D(728, (3,3), padding='same')(x)
x = BatchNormalization()(x)
x = add7 = Add()([add6, x])
#BLOCK 9
x = Activation('relu')(x)
x = SeparableConv2D(728, (3,3), padding='same')(x)
x = BatchNormalization()(x)
x = concatenate([conv6, x], axis = 3)
x = Activation('relu')(x)
x = SeparableConv2D(728, (3,3), padding='same')(x)
x = BatchNormalization()(x)
x = Activation('relu')(x)
x = SeparableConv2D(728, (3,3), padding='same')(x)
x = BatchNormalization()(x)
x = add8 = Add()([add7, x])
#BLOCK 10
x = Activation('relu')(x)
x = SeparableConv2D(728, (3,3), padding='same')(x)
x = BatchNormalization()(x)
x = concatenate([conv5, x], axis = 3)
x = Activation('relu')(x)
x = SeparableConv2D(728, (3,3), padding='same', name ='blockmiddle_end')(x)
x = BatchNormalization()(x)
x = Activation('relu')(x)
x = SeparableConv2D(728, (3,3), padding='same')(x)
x = BatchNormalization()(x)
x = add9 = Add()([add8, x])
#BLOCK 11
x = UpSampling2D((2,2))(x)
x = Activation('relu')(x)
x = SeparableConv2D(728, (3,3), padding='same', name='block4_start')(x)
x = BatchNormalization()(x)
x = concatenate([conv4, x], axis = 3)
x = Activation('relu')(x)
x = SeparableConv2D(728, (3,3), padding='same')(x)
x = BatchNormalization()(x)
c45 = Conv2D(728, (1,1))(UpSampling2D()(add9))
x = add10 = Add()([c45, x])
#BLOCK 12
x = Conv2DTranspose(1, (3,3), strides=(2,2))(x)
x = Activation('relu')(x)
x = SeparableConv2D(256, (3,3), padding='valid', name='block3_start')(x)
x = BatchNormalization()(x)
x = concatenate([conv3, x], axis = 3)
x = Activation('relu')(x)
x = SeparableConv2D(256, (3,3), padding='same')(x)
x = BatchNormalization()(x)
c34 = Conv2D(256, (3,3), padding='valid')(Conv2DTranspose(1, (3,3), strides=(2,2))(add10))
x = add11 = Add()([c34, x])
#BLOCK 13
x = Conv2DTranspose(1, (3,3), strides=(2,2))(x)
x = Activation('relu')(x)
x = SeparableConv2D(128, (3,3), padding='valid', name='block2_start')(x)
x = BatchNormalization()(x)
x = concatenate([conv2, x], axis = 3)
x = Activation('relu')(x)
x = SeparableConv2D(128, (3,3), padding='same')(x)
x = BatchNormalization()(x)
c23 = Conv2D(128, (3,3), padding='valid')(Conv2DTranspose(1, (3,3), strides=(2,2))(add11))
x = add12 = Add()([c23, x])
#BLOCK 14
x = Conv2D(64, (3,3), padding='same', name='block1_start')(x)
x = BatchNormalization()(x)
x = Activation('relu')(x)
x = concatenate([conv1, x], axis = 3)
x = ZeroPadding2D()(x)
x = Conv2D(32, (3,3), padding='same')(x)
x = BatchNormalization()(x)
x = Activation('relu')(x)
x = UpSampling2D()(x)
x = ZeroPadding2D()(x)
x = Conv2D(2, 3, activation = 'relu', padding = 'same',)(x)
mask = x = Conv2D(3, 1, activation = 'sigmoid',name='Mask')(x)
base_model2 = tf.keras.applications.Xception(include_top=False, weights='imagenet',input_shape = (224,224,3))
x2_0 = base_model2(mask)
x2_0 = Flatten(name='Flatten2')(x2_0)
x2_1 = Dense(256, name='fc2',activation='relu')(x2_0)
classif_out_encoder2 = Dense(nbr_of_classes, name='out2',activation='softmax')(x2_1)
#Create ResTS Model
model = Model(base_model1.input, [classif_out_encoder1, classif_out_encoder2])
#Compile the mode to use multi-task learning
losses = {
"out1": 'categorical_crossentropy',
"out2": 'categorical_crossentropy'
}
alpha=0.4
lossWeights = {"out1": alpha, "out2": (1.0-alpha)}
model.compile(optimizer=optimizers.SGD(lr=1e-4, momentum=0.9), loss=losses, loss_weights=lossWeights, metrics = ['accuracy'])
model.summary()
nb_epoch = 15
history = model.fit(train(),
steps_per_epoch = 2545,
epochs = nb_epoch,
validation_data=valid(),
validation_steps = 678)
df = pd.DataFrame(history.history)
df.to_csv('ResTS15epochs.csv')
try:
model.save('ResTS.h5')
except:
print('Check if the model has been saved!')