-
Notifications
You must be signed in to change notification settings - Fork 0
/
RexNet.py
475 lines (446 loc) · 14.7 KB
/
RexNet.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
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
import copy
import tensorflow.python.keras.layers as layers
from keras import backend as K
from keras.layers import Dense, GlobalAveragePooling2D, Multiply
from tensorflow import keras
from tensorflow.keras.models import Model
from tensorflow.python.keras import activations, backend
from tensorflow.python.keras.applications import imagenet_utils
from tensorflow.python.keras.engine import training
from tensorflow.python.keras.utils import data_utils, layer_utils
DEFAULT_BLOCKS_ARGS = [{
'filters_in': 32,
'filters_out': 16,
'expand_ratio': 1,
'bn_momentum': 0.9,
'drop_ratio': 0.2,
'use_se': False,
'se_ratio': 0,
}, {
'filters_in': 16,
'filters_out': 27,
'expand_ratio': 6,
'bn_momentum': 0.9,
'drop_ratio': 0.2,
'use_se': False,
'se_ratio': 0,
}, {
'filters_in': 27,
'filters_out': 38,
'expand_ratio': 6,
'bn_momentum': 0.9,
'drop_ratio': 0.2,
'use_se': False,
'se_ratio': 0,
}, {
'filters_in': 38,
'filters_out': 50,
'expand_ratio': 6,
'bn_momentum': 0.9,
'drop_ratio': 0.2,
'use_se': True,
'se_ratio': 12,
}, {
'filters_in': 50,
'filters_out': 61,
'expand_ratio': 6,
'bn_momentum': 0.9,
'drop_ratio': 0.2,
'use_se': True,
'se_ratio': 12,
}, {
'filters_in': 61,
'filters_out': 72,
'expand_ratio': 6,
'bn_momentum': 0.9,
'drop_ratio': 0.2,
'use_se': True,
'se_ratio': 12,
}, {
'filters_in': 72,
'filters_out': 84,
'expand_ratio': 6,
'bn_momentum': 0.9,
'drop_ratio': 0.2,
'use_se': True,
'se_ratio': 12,
}, {
'filters_in': 84,
'filters_out': 95,
'expand_ratio': 6,
'bn_momentum': 0.9,
'drop_ratio': 0.2,
'use_se': True,
'se_ratio': 12,
}, {
'filters_in': 95,
'filters_out': 106,
'expand_ratio': 6,
'bn_momentum': 0.9,
'drop_ratio': 0.2,
'use_se': True,
'se_ratio': 12,
}, {
'filters_in': 106,
'filters_out': 117,
'expand_ratio': 6,
'bn_momentum': 0.9,
'drop_ratio': 0.2,
'use_se': True,
'se_ratio': 12,
}, {
'filters_in': 117,
'filters_out': 120,
'expand_ratio': 6,
'bn_momentum': 0.9,
'drop_ratio': 0.2,
'use_se': True,
'se_ratio': 12,
}, {
'filters_in': 128,
'filters_out': 140,
'expand_ratio': 6,
'bn_momentum': 0.9,
'drop_ratio': 0.2,
'use_se': True,
'se_ratio': 12,
}, {
'filters_in': 140,
'filters_out': 151,
'expand_ratio': 6,
'bn_momentum': 0.9,
'drop_ratio': 0.2,
'use_se': True,
'se_ratio': 12,
}, {
'filters_in': 151,
'filters_out': 162,
'expand_ratio': 6,
'bn_momentum': 0.9,
'drop_ratio': 0.2,
'use_se': True,
'se_ratio': 12,
}, {
'filters_in': 162,
'filters_out': 174,
'expand_ratio': 6,
'bn_momentum': 0.9,
'drop_ratio': 0.2,
'use_se': True,
'se_ratio': 12,
}, {
'filters_in': 174,
'filters_out': 185,
'expand_ratio': 6,
'bn_momentum': 0.9,
'drop_ratio': 0.2,
'use_se': True,
'se_ratio': 12,
}]
def ReXNet(
input_tensor=None,
input_shape=(224, 224, 3),
activation="swish",
use_bias=False,
alpha=1.0,
se_ratio=4,
blocks_args='default',
include_top=True,
weights=None,
classes=4,
classifier_activation='softmax',
default_size=224,
bn_momentum=0.9,
pooling='avg'
):
""" ReXNetV1 architecture.
Reference:
- [ReXNet: Diminishing Representational Bottleneck on Convolutional Neural //
Network](
https://arxiv.org/pdf/2007.00992v1.pdf) (CVPR 2020)
Optionally loads weights pre-trained on ImageNet.
Arguments:
input_shape: optional shape tuple, only to be specified
if `include_top` is False.
It should have exactly 3 inputs channels.
input_tensor: optional Keras tensor
(i.e. output of `layers.Input()`)
to use as image input for the model.
activation: A `str` or callable. The activation function to use
on the "final" layer.
use_bias : Boolean, whether the layer uses a bias vector.
alpha: Float between 0 and 1. controls the width of the network.
This is known as the width multiplier in the ReXnet paper .
- If `alpha` < 1.0, proportionally decreases the number
of filters in each layer.
- If `alpha` > 1.0, proportionally increases the number
of filters in each layer.
- If `alpha` = 1, default number of filters from the paper
are used at each layer.
se_ratio: float between 0 and 1, fraction to squeeze the input filters.
include_top: whether to include the fully-connected
layer at the top of the network.
weights: one of `None` (random initialization),
'imagenet' (pre-training on ImageNet),
or the path to the weights file to be loaded.
classes: Integer, optional number of classes to classify images
into, only to be specified if `include_top` is True, and
if no `weights` argument is specified.
default_size: integer, default input image size.
Returns:
A `keras.Model` .
Raises:
ValueError: in case of invalid argument for `weights`, or invalid input //
shape.
"""
input_shape = imagenet_utils.obtain_input_shape(
input_shape,
default_size=default_size,
min_size=32,
data_format=backend.image_data_format(),
require_flatten=include_top,
weights=None,
)
if blocks_args == 'default':
blocks_args = DEFAULT_BLOCKS_ARGS
if input_tensor is None:
img_input = layers.Input(shape=input_shape)
else:
if not backend.is_keras_tensor(input_tensor):
img_input = layers.Input(tensor=input_tensor, shape=input_shape)
else:
img_input = input_tensor
channel_axis = 1 if backend.image_data_format() == "channels_first" else -1
# ReXNet architecture :
# stem layer : (conv 3x3 , 2 stride , swish activation)
# conBSwish_1 = conv2d_BN(
# img_input, 32, 3, (2, 2), padding="valid", use_bias=use_bias ,bn_momentum= bn_momentum
# )
conv = layers.Conv2D(
32,
3,
strides=2,
padding='valid',
use_bias=False,
name='stem_conv')(img_input)
convbn = layers.BatchNormalization(axis=channel_axis,
momentum=bn_momentum, name='stem_bn')(conv)
conBNswish = layers.Activation(activation, name='stem_activation')(convbn)
# inverted_Bottleneck_block : (Expand + Depthwise + Squeeze Excitation //
# Module (optional) + Project )
blocks_args = copy.deepcopy(blocks_args)
x = conBNswish
for (i, args) in enumerate(blocks_args):
x = inverted_residual_block(
x,
alpha=alpha,
name='block_{}_'.format(i + 1),
**args)
# penultimate layer : (conv 1X1 , 1 stride , swish activation )
# AveragePooling2D + Fullyconnected
pen_channels = 1280
# block_17 = layers.Dropout(0.2)(block_16)
conBNswish_2 = conv2d_BN(x, 1280, 1, (1, 1), bn_momentum=bn_momentum)
Average_Pooling_layer = layers.AveragePooling2D((1, 1), padding="same")(
conBNswish_2
)
block_17 = layers.Dropout(0.2)(Average_Pooling_layer)
# # FC_layer = conv2d_BN(block_17, 1280, 1, (1, 1), name="conv3")
FC_layer = layers.Conv2D(
4,
kernel_size=1,
strides=1, use_bias=True,
)(block_17)
# Ensure that the model takes into account any potential predecessors \\
# of `input_tensor`.
# FC_layer = layers.Flatten()(FC_layer)
# x = layers.GlobalAveragePooling2D()(FC_layer)
# x = layers.Dense(classes, activation=classifier_activation,
# name='predictions')(x)
x = layers.GlobalAveragePooling2D()(FC_layer)
# if pooling == 'avg':
# x = layers.GlobalAveragePooling2D()(FC_layer)
# elif pooling == 'max':
# x = layers.GlobalMaxPooling2D()(FC_layer)
if input_tensor is not None:
inputs = layer_utils.get_source_inputs(input_tensor)
else:
inputs = img_input
# create the model
model = training.Model(inputs, x, name="ReXNet_V1")
return model
def conv2d_BN(
input_tensor,
filter,
kernel_size,
strides,
bn_momentum,
padding="same",
activation="swish",
use_bias=False,
name="top_conv",
):
"""A block that has a conv layer at shortcut.
Arguments :
input_tensor: input tensor
filters: Integer, the dimensionality of the output space //
(i.e. the number of output filters in the convolution).
kernel_size: An integer or tuple/list of 2 integers, specifying the //
height and width of the 2D convolution window.
strides: An integer or tuple/list of 2 integers, specifying the //
strides of the convolution along the height and width.
padding : one of "valid" or "same"
activation : Activation function to use. If you don't specify anything,//
no activation is applied (see keras.activations).
use_bias : Boolean, whether the layer uses a bias vector.
name : layer name .
Returns :
Output tensor for the block.
"""
input_tensor = layers.Conv2D(
filter,
kernel_size,
strides=strides,
padding=padding,
use_bias=use_bias,
name=name,
)(input_tensor)
if not use_bias:
bn_axis = 1 if backend.image_data_format() == "channels_first" else -1
bn_name = None if name is None else name + "_bn"
input_tensor = layers.BatchNormalization(
axis=bn_axis, epsilon=1e-05, momentum=bn_momentum,
scale=False, name=bn_name
)(input_tensor)
if activation is not None:
an_name = None if name is None else name + "_ac"
input_tensor = layers.Activation(activation, name=an_name)(
input_tensor
)
return input_tensor
def inverted_residual_block(inputs,
alpha,
name='',
filters_in=32,
filters_out=16,
expand_ratio=1,
bn_momentum=0.9,
drop_ratio=0.2,
use_se=False,
se_ratio=0.):
"""An inverted residual block.
Arguments:
inputs: input tensor.
drop_rate: float between 0 and 1, fraction of the input units to drop.
name: string, block label.
filters_in: integer, the number of input filters.
filters_out: integer, the number of output filters.
kernel_size: integer, the dimension of the convolution window.
expand_ratio: integer, scaling coefficient for the input filters.
se_ratio: float between 0 and 1, fraction to squeeze the input filters.
use_se : boolean , for using squeeze and excitation block.
Returns:
output tensor for the block.
"""
bn_axis = 3 if backend.image_data_format() == 'channels_last' else 1
channel = int(filters_out * alpha)
# Expansion phase
filters = filters_in * expand_ratio
if expand_ratio != 1:
x = layers.Conv2D(
filters,
kernel_size=1,
strides=1,
padding='same',
use_bias=False,
name=name + 'expand_conv')(
inputs)
x = layers.BatchNormalization(axis=bn_axis, momentum=bn_momentum
, name=name + 'expand_bn')(x)
x = layers.Activation("swish", name=name + 'expand_activation')(x)
else:
x = inputs
# Depthwise Convolution
conv_pad = 'same'
x = layers.DepthwiseConv2D(
kernel_size=3,
strides=1,
padding=conv_pad,
use_bias=False,
name=name + 'dwconv')(x)
x = layers.BatchNormalization(axis=bn_axis, momentum=bn_momentum
, name=name + 'bn')(x)
x = layers.Activation('relu', name=name + 'activation')(x)
# Squeeze and Excitation phase
if use_se:
# filters_se = max(1, int(filters_in * se_ratio))
nb_chan = K.int_shape(x)[-1]
filters_se = nb_chan // se_ratio
se = layers.GlobalAveragePooling2D(name=name + 'se_squeeze')(x)
se = layers.Reshape((1, 1, filters), name=name + 'se_reshape')(se)
se = layers.Conv2D(
filters_se,
1,
padding='same',
activation='relu',
name=name + 'se_reduce')(
se)
se = layers.Conv2D(
filters,
1,
padding='same',
activation='sigmoid',
name=name + 'se_expand')(se)
x = layers.multiply([x, se], name=name + 'se_excite')
# Output phase
x = layers.Conv2D(
channel,
1,
padding='same',
use_bias=False,
name=name + 'project_conv')(x)
x = layers.BatchNormalization(axis=bn_axis, momentum=bn_momentum,
name=name + 'project_bn')(x)
if filters_in <= filters_out:
y = layers.Conv2D(filters=filters_out, kernel_size=1)(inputs)
y = layers.BatchNormalization(axis=bn_axis)(y)
x += y
return x
def ReXNetV1(ReXNetV1_config, task, train_config):
"""
Rexnetv1 final architetcure add last layer according to the task in the //
input .
Arguments :
ReXNetV1_config : ReXNetV1 config file
task : string , regression / classification
returns :
A `keras.Model` .
"""
ReXNetV1 = ReXNet(include_top=ReXNetV1_config.INCLUDE_TOP,
weights=ReXNetV1_config.WEIGHTS,
input_tensor=None,
pooling=ReXNetV1_config.POOLING,
input_shape=(
ReXNetV1_config.INPUT_SHAPE.HEIGHT,
ReXNetV1_config.INPUT_SHAPE.WIDTH,
ReXNetV1_config.INPUT_SHAPE.CHANNELS
)
)
base_model = ReXNetV1.output
if ReXNetV1_config.DROPOUT.BOOL == True:
base_model = layers.Dropout(ReXNetV1_config.DROPOUT.PROB)(base_model)
if task == "regression":
predection = layers.Dense(1, activation="linear")(base_model)
elif task == "binary_classification" or task == "multiclass_classification":
if train_config.ACTIVATION == 'sigmoid':
predection = layers.Dense(1, activation="sigmoid")(base_model)
elif train_config.ACTIVATION == 'softmax':
predection = layers.Dense(
train_config.CLASSES, activation="softmax"
)(base_model)
else:
raise Exception("none of the activation functions are true !")
else:
raise Exception("task not available!")
model = Model(inputs=ReXNetV1.input, outputs=predection)
return model