-
Notifications
You must be signed in to change notification settings - Fork 1
/
Copy pathdeeplab_v3_plus.py
296 lines (252 loc) · 13.8 KB
/
deeplab_v3_plus.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
import os
import numpy as np
import tensorflow as tf
from keras.models import Model # Functional API
from keras import layers
from keras.layers import (
Conv2D, Input, Dropout, Lambda, Concatenate,
BatchNormalization, Activation, AveragePooling2D,
ZeroPadding2D, DepthwiseConv2D, Add)
from keras.optimizers import Adam
from keras.activations import relu
#from tensorflow.keras import backend as K
from src.metrics import (dice, jaccard)
from src.engine import scale_input
from src.utils import make_divisible
# Deeplab-v3+
class DSC_DeepLab_v3_plus:
def __init__(self,
pre_trained=False, # if True, set weights_path
weights_path=None, # full-path to the pre-trained models_weights
backbone='mobilenetv2',
n_classes=None,
input_h=None,
input_w=None,
activation='relu',
kernel_init='he_normal',
model_name=None
):
self.pre_trained = pre_trained
self.weights_path = weights_path
self.backbone = backbone
self.n_classes = n_classes
self.input_h = input_h
self.input_w = input_w
self.activation = activation
self.kernel_init = kernel_init
self.model_name = model_name
self.OS = 16
self.alpha = 1.
def depth_sep_conv(self, x, filters, stride=1, kernel_size=3, rate=1, depth_activation=False, eps=1e-3):
"""
Depth-wise Convolution:
1. Depth-wise Separable Convolution
2. Point-wise Convolution
"""
if stride == 1:
depth_padding = 'same'
else:
kernel_size_effective = kernel_size + (kernel_size - 1) * (rate - 1)
pad_total = kernel_size_effective - 1
pad_beg = pad_total // 2
pad_end = pad_total - pad_beg
x = ZeroPadding2D((pad_beg, pad_end))(x)
depth_padding = 'valid'
if not depth_activation:
x = Activation(self.activation)(x)
x = DepthwiseConv2D((kernel_size, kernel_size), strides=(stride, stride), dilation_rate=(rate, rate), padding=depth_padding, use_bias=False)(x)
x = BatchNormalization(epsilon=eps)(x)
if depth_activation:
x = Activation(self.activation)(x)
x = Conv2D(filters, (1, 1), padding='same', use_bias=False)(x)
x = BatchNormalization(epsilon=eps)(x)
if depth_activation:
x = Activation(self.activation)(x)
return x
def conv2d_same(self, x, filters, stride=1, kernel_size=3, rate=1):
if stride == 1:
return Conv2D(filters,
(kernel_size, kernel_size),
strides=(stride, stride),
padding='same', use_bias=False,
dilation_rate=(rate, rate))(x)
else:
kernel_size_effective = kernel_size + (kernel_size - 1) * (rate - 1)
pad_total = kernel_size_effective - 1
pad_beg = pad_total // 2
pad_end = pad_total - pad_beg
x = ZeroPadding2D((pad_beg, pad_end))(x)
return Conv2D(filters,
(kernel_size, kernel_size),
strides=(stride, stride),
padding='valid', use_bias=False,
dilation_rate=(rate, rate))(x)
def xception_block(self, inputs, depth_list, skip_connect_type, stride, rate=1, depth_activation=False, return_skip=False):
residual = inputs
for i in range(3):
residual = self.depth_sep_conv(residual,
depth_list[i],
stride=stride if i == 2 else 1,
rate=rate,
depth_activation=depth_activation)
if i == 1:
skip = residual
if skip_connect_type == 'conv':
shortcut = self.conv2d_same(inputs, depth_list[-1],
kernel_size=1,
stride=stride)
shortcut = BatchNormalization()(shortcut)
outputs = layers.add([residual, shortcut])
elif skip_connect_type == 'sum':
outputs = layers.add([residual, inputs])
elif skip_connect_type == 'none':
outputs = residual
if return_skip:
return outputs, skip
else:
return outputs
def inverted_res_block(self, inputs, expansion, stride, alpha, filters, block_id, skip_connection, rate=1, eps=1e-3):
in_channels = inputs._keras_shape[-1]
pointwise_conv_filters = int(filters * alpha)
pointwise_filters = make_divisible(pointwise_conv_filters, 8)
x = inputs
if block_id:
# Expand
x = Conv2D(expansion * in_channels, kernel_size=1, padding='same',
use_bias=False, activation=None)(x)
x = BatchNormalization(epsilon=1e-3, momentum=0.999)(x)
x = Lambda(lambda x: relu(x, max_value=6.))(x)
# Depthwise
x = DepthwiseConv2D(kernel_size=3, strides=stride, activation=None, use_bias=False, padding='same', dilation_rate=(rate, rate))(x)
x = BatchNormalization(epsilon=1e-3, momentum=0.999)(x)
x = Lambda(lambda x: relu(x, max_value=6.))(x)
x = Conv2D(pointwise_filters,kernel_size=1, padding='same', use_bias=False, activation=None)(x)
x = BatchNormalization(epsilon=1e-3, momentum=0.999)(x)
if skip_connection:
return Add()([inputs, x])
return x
def build(self):
if not (self.backbone in {'xception', 'mobilenetv2'}):
raise ValueError('The `backbone` argument should be either `xception` or `mobilenetv2` ')
img_input = Input(shape=(self.input_h, self.input_w, 3))
# Lambda layer: scale input before feeding to the network
batches_input = Lambda(lambda x: scale_input(x))(img_input)
if self.backbone == 'xception':
if self.OS == 8:
entry_block3_stride = 1
middle_block_rate = 2 # ! Not mentioned in paper, but required
exit_block_rates = (2, 4)
atrous_rates = (12, 24, 36)
else:
entry_block3_stride = 2
middle_block_rate = 1
exit_block_rates = (1, 2)
atrous_rates = (6, 12, 18)
x = Conv2D(32, (3, 3), strides=(2, 2), use_bias=False, padding='same')(batches_input)
x = BatchNormalization(name='entry_flow_conv1_1_BN')(x)
x = Activation(self.activation)(x)
x = self.conv2d_same(x, 64, kernel_size=3, stride=1)
x = BatchNormalization()(x)
x = Activation(self.activation)(x)
x = self.xception_block(x, [128, 128, 128], skip_connect_type='conv', stride=2, depth_activation=False)
x, skip1 = self.xception_block(x, [256, 256, 256], skip_connect_type='conv', stride=2, depth_activation=False, return_skip=True)
x = self.xception_block(x, [728, 728, 728], skip_connect_type='conv', stride=entry_block3_stride, depth_activation=False)
for i in range(16):
x = self.xception_block(x, [728, 728, 728], skip_connect_type='sum', stride=1, rate=middle_block_rate, depth_activation=False)
x = self.xception_block(x, [728, 1024, 1024], skip_connect_type='conv', stride=1, rate=exit_block_rates[0], depth_activation=False)
x = self.xception_block(x, [1536, 1536, 2048], skip_connect_type='none', stride=1, rate=exit_block_rates[1], depth_activation=True)
# Backbone='mobilenetv2'
else:
self.OS = 8
first_block_filters = make_divisible(32 * self.alpha, 8)
x = Conv2D(first_block_filters, kernel_size=3, strides=(2, 2), padding='same', use_bias=False)(batches_input)
x = BatchNormalization(epsilon=1e-3, momentum=0.999)(x)
x = Lambda(lambda x: relu(x, max_value=6.))(x)
x = self.inverted_res_block(x, filters=16, alpha=self.alpha, stride=1, expansion=1, block_id=0, skip_connection=False)
x = self.inverted_res_block(x, filters=24, alpha=self.alpha, stride=2,
expansion=6, block_id=1, skip_connection=False)
x = self.inverted_res_block(x, filters=24, alpha=self.alpha, stride=1,
expansion=6, block_id=2, skip_connection=True)
x = self.inverted_res_block(x, filters=32, alpha=self.alpha, stride=2,
expansion=6, block_id=3, skip_connection=False)
x = self.inverted_res_block(x, filters=32, alpha=self.alpha, stride=1,
expansion=6, block_id=4, skip_connection=True)
x = self.inverted_res_block(x, filters=32, alpha=self.alpha, stride=1,
expansion=6, block_id=5, skip_connection=True)
# stride in block 6 changed from 2 -> 1, so we need to use rate = 2
x = self.inverted_res_block(x, filters=64, alpha=self.alpha, stride=1, # 1!
expansion=6, block_id=6, skip_connection=False)
x = self.inverted_res_block(x, filters=64, alpha=self.alpha, stride=1, rate=2,
expansion=6, block_id=7, skip_connection=True)
x = self.inverted_res_block(x, filters=64, alpha=self.alpha, stride=1, rate=2,
expansion=6, block_id=8, skip_connection=True)
x = self.inverted_res_block(x, filters=64, alpha=self.alpha, stride=1, rate=2,
expansion=6, block_id=9, skip_connection=True)
x = self.inverted_res_block(x, filters=96, alpha=self.alpha, stride=1, rate=2,
expansion=6, block_id=10, skip_connection=False)
x = self.inverted_res_block(x, filters=96, alpha=self.alpha, stride=1, rate=2,
expansion=6, block_id=11, skip_connection=True)
x = self.inverted_res_block(x, filters=96, alpha=self.alpha, stride=1, rate=2,
expansion=6, block_id=12, skip_connection=True)
x = self.inverted_res_block(x, filters=160, alpha=self.alpha, stride=1, rate=2, # 1!
expansion=6, block_id=13, skip_connection=False)
x = self.inverted_res_block(x, filters=160, alpha=self.alpha, stride=1, rate=4,
expansion=6, block_id=14, skip_connection=True)
x = self.inverted_res_block(x, filters=160, alpha=self.alpha, stride=1, rate=4,
expansion=6, block_id=15, skip_connection=True)
x = self.inverted_res_block(x, filters=320, alpha=self.alpha, stride=1, rate=4,
expansion=6, block_id=16, skip_connection=False)
# Image Feature branch
b4 = AveragePooling2D(pool_size=(int(np.ceil(self.input_h / self.OS)), int(np.ceil(self.input_w / self.OS))))(x)
b4 = Conv2D(256, (1, 1), padding='same', use_bias=False, name='image_pooling')(b4)
b4 = BatchNormalization(epsilon=1e-5)(b4)
b4 = Activation(self.activation)(b4)
b4 = Lambda(lambda x: tf.compat.v1.image.resize_bilinear(x, size=(int(np.ceil(self.input_h / self.OS)), int(np.ceil(self.input_w / self.OS)))))(b4)
# simple 1x1
b0 = Conv2D(256, (1, 1), padding='same', use_bias=False, name='aspp0')(x)
b0 = BatchNormalization(epsilon=1e-5)(b0)
b0 = Activation(self.activation)(b0)
# there are only 2 branches in mobilenetV2. not sure why
if self.backbone == 'xception':
# rate = 6 (12)
b1 = self.depth_sep_conv(x, 256, rate=atrous_rates[0], depth_activation=True, eps=1e-5)
# rate = 12 (24)
b2 = self.depth_sep_conv(x, 256, rate=atrous_rates[1], depth_activation=True, eps=1e-5)
# rate = 18 (36)
b3 = self.depth_sep_conv(x, 256, rate=atrous_rates[2], depth_activation=True, eps=1e-5)
# concatenate ASPP branches & project
x = Concatenate()([b4, b0, b1, b2, b3])
else:
x = Concatenate()([b4, b0])
x = Conv2D(256, (1, 1), padding='same', use_bias=False)(x)
x = BatchNormalization(epsilon=1e-5)(x)
x = Activation(self.activation)(x)
x = Dropout(0.1)(x)
# DeepLab v.3+ decoder
if self.backbone == 'xception':
# Feature projection
# x4 (x2) block
x = Lambda(lambda x: tf.compat.v1.image.resize_bilinear(x, size=(int(np.ceil(self.input_h / 4)), int(np.ceil(self.input_w / 4)))))(x)
dec_skip1 = Conv2D(48, (1, 1), padding='same', use_bias=False)(skip1)
dec_skip1 = BatchNormalization(epsilon=1e-5)(dec_skip1)
dec_skip1 = Activation(self.activation)(dec_skip1)
x = Concatenate()([x, dec_skip1])
x = self.depth_sep_conv(x, 256, depth_activation=True, eps=1e-5)
x = self.depth_sep_conv(x, 256, depth_activation=True, eps=1e-5)
# Output layer
x = Conv2D(self.n_classes, (1, 1), padding='same')(x)
x = Lambda(lambda x: tf.compat.v1.image.resize_bilinear(x, size=(self.input_h, self.input_w)))(x)
x = Activation('softmax')(x)
# Create model
model = Model(inputs=img_input, outputs=x)
model.compile(optimizer=Adam(),
loss="categorical_crossentropy",
metrics=[dice, jaccard, ]
)
# Load models_weights if pre-trained
if self.pre_trained:
if os.path.exists(self.weights_path):
model.load_weights(self.weights_path)
else:
raise Exception(f'Failed to load weights {self.weights_path}')
return model