forked from jinuvision/autocv
-
Notifications
You must be signed in to change notification settings - Fork 0
/
mobilenet_v2.py
329 lines (307 loc) · 13.4 KB
/
mobilenet_v2.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
# Copyright 2018 The TensorFlow Authors. All Rights Reserved.
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
# http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.
# ==============================================================================
# Modified by: Heung-Chang Lee, Do-Guk Kim
"""Implementation of Mobilenet V2.
Architecture: https://arxiv.org/abs/1801.04381
The base model gives 72.2% accuracy on ImageNet, with 300MMadds,
3.4 M parameters.
"""
from __future__ import absolute_import
from __future__ import division
from __future__ import print_function
import copy
import functools
import tensorflow as tf
import conv_blocks as ops
import mobilenet as lib
slim = tf.contrib.slim
op = lib.op
expand_input = ops.expand_input_by_factor
@slim.add_arg_scope
def mobilenet(input_tensor,
input_size,
num_classes=1001,
depth_multiplier=1.0,
scope='MobilenetV2',
conv_defs=None,
finegrain_classification_mode=False,
min_depth=None,
divisible_by=None,
activation_fn=None,
**kwargs):
"""Creates mobilenet V2 network.
Inference mode is created by default. To create training use training_scope
below.
with tf.contrib.slim.arg_scope(mobilenet_v2.training_scope()):
logits, endpoints = mobilenet_v2.mobilenet(input_tensor)
Args:
input_tensor: The input tensor
num_classes: number of classes
depth_multiplier: The multiplier applied to scale number of
channels in each layer.
scope: Scope of the operator
conv_defs: Allows to override default conv def.
finegrain_classification_mode: When set to True, the model
will keep the last layer large even for small multipliers. Following
https://arxiv.org/abs/1801.04381
suggests that it improves performance for ImageNet-type of problems.
*Note* ignored if final_endpoint makes the builder exit earlier.
min_depth: If provided, will ensure that all layers will have that
many channels after application of depth multiplier.
divisible_by: If provided will ensure that all layers # channels
will be divisible by this number.
activation_fn: Activation function to use, defaults to tf.nn.relu6 if not
specified.
**kwargs: passed directly to mobilenet.mobilenet:
prediction_fn- what prediction function to use.
reuse-: whether to reuse variables (if reuse set to true, scope
must be given).
Returns:
logits/endpoints pair
Raises:
ValueError: On invalid arguments
"""
global V2_DEF
if input_size > 168:
V2_DEF = dict(
defaults={
# Note: these parameters of batch norm affect the architecture
# that's why they are here and not in training_scope.
(slim.batch_norm,): {'center': True, 'scale': True},
(slim.conv2d, slim.fully_connected, slim.separable_conv2d): {
'normalizer_fn': slim.batch_norm, 'activation_fn': tf.nn.relu6
},
(ops.expanded_conv,): {
'expansion_size': expand_input(6),
'split_expansion': 1,
'normalizer_fn': slim.batch_norm,
'residual': True
},
(slim.conv2d, slim.separable_conv2d): {'padding': 'SAME'}
},
spec=[
op(slim.conv2d, stride=2, num_outputs=32, kernel_size=[3, 3]),
op(ops.expanded_conv,
expansion_size=expand_input(1, divisible_by=1),
num_outputs=16),
op(ops.expanded_conv, stride=2, num_outputs=24),
op(ops.expanded_conv, stride=1, num_outputs=24),
op(ops.expanded_conv, stride=2, num_outputs=32),
op(ops.expanded_conv, stride=1, num_outputs=32),
op(ops.expanded_conv, stride=1, num_outputs=32),
op(ops.expanded_conv, stride=2, num_outputs=64),
op(ops.expanded_conv, stride=1, num_outputs=64),
op(ops.expanded_conv, stride=1, num_outputs=64),
op(ops.expanded_conv, stride=1, num_outputs=64),
op(ops.expanded_conv, stride=1, num_outputs=96),
op(ops.expanded_conv, stride=1, num_outputs=96),
op(ops.expanded_conv, stride=1, num_outputs=96),
op(ops.expanded_conv, stride=2, num_outputs=160),
op(ops.expanded_conv, stride=1, num_outputs=160),
# op(ops.expanded_conv, stride=1, num_outputs=160),
# op(ops.expanded_conv, stride=1, num_outputs=320),
# op(slim.conv2d, stride=1, kernel_size=[1, 1], num_outputs=1280)
],
)
elif input_size > 84:
V2_DEF = dict(
defaults={
# Note: these parameters of batch norm affect the architecture
# that's why they are here and not in training_scope.
(slim.batch_norm,): {'center': True, 'scale': True},
(slim.conv2d, slim.fully_connected, slim.separable_conv2d): {
'normalizer_fn': slim.batch_norm, 'activation_fn': tf.nn.relu6
},
(ops.expanded_conv,): {
'expansion_size': expand_input(6),
'split_expansion': 1,
'normalizer_fn': slim.batch_norm,
'residual': True
},
(slim.conv2d, slim.separable_conv2d): {'padding': 'SAME'}
},
spec=[
op(slim.conv2d, stride=2, num_outputs=32, kernel_size=[3, 3]),
op(ops.expanded_conv,
expansion_size=expand_input(1, divisible_by=1),
num_outputs=16),
op(ops.expanded_conv, stride=2, num_outputs=24),
op(ops.expanded_conv, stride=1, num_outputs=24),
op(ops.expanded_conv, stride=2, num_outputs=32),
op(ops.expanded_conv, stride=1, num_outputs=32),
op(ops.expanded_conv, stride=1, num_outputs=32),
op(ops.expanded_conv, stride=2, num_outputs=64),
op(ops.expanded_conv, stride=1, num_outputs=64),
# op(ops.expanded_conv, stride=1, num_outputs=64),
# op(ops.expanded_conv, stride=1, num_outputs=64),
# op(ops.expanded_conv, stride=1, num_outputs=96),
# op(ops.expanded_conv, stride=1, num_outputs=96),
# op(ops.expanded_conv, stride=1, num_outputs=96),
# op(ops.expanded_conv, stride=2, num_outputs=160),
# op(ops.expanded_conv, stride=1, num_outputs=160),
# op(ops.expanded_conv, stride=1, num_outputs=160),
# op(ops.expanded_conv, stride=1, num_outputs=320),
# op(slim.conv2d, stride=1, kernel_size=[1, 1], num_outputs=1280)
],
)
elif input_size > 28:
V2_DEF = dict(
defaults={
# Note: these parameters of batch norm affect the architecture
# that's why they are here and not in training_scope.
(slim.batch_norm,): {'center': True, 'scale': True},
(slim.conv2d, slim.fully_connected, slim.separable_conv2d): {
'normalizer_fn': slim.batch_norm, 'activation_fn': tf.nn.relu6
},
(ops.expanded_conv,): {
'expansion_size': expand_input(6),
'split_expansion': 1,
'normalizer_fn': slim.batch_norm,
'residual': True
},
(slim.conv2d, slim.separable_conv2d): {'padding': 'SAME'}
},
spec=[
op(slim.conv2d, stride=2, num_outputs=32, kernel_size=[3, 3]),
op(ops.expanded_conv,
expansion_size=expand_input(1, divisible_by=1),
num_outputs=16),
op(ops.expanded_conv, stride=2, num_outputs=24),
op(ops.expanded_conv, stride=1, num_outputs=24),
op(ops.expanded_conv, stride=2, num_outputs=32),
op(ops.expanded_conv, stride=1, num_outputs=32),
# op(ops.expanded_conv, stride=1, num_outputs=32),
# op(ops.expanded_conv, stride=2, num_outputs=64),
# op(ops.expanded_conv, stride=1, num_outputs=64),
# op(ops.expanded_conv, stride=1, num_outputs=64),
# op(ops.expanded_conv, stride=1, num_outputs=64),
# op(ops.expanded_conv, stride=1, num_outputs=96),
# op(ops.expanded_conv, stride=1, num_outputs=96),
# op(ops.expanded_conv, stride=1, num_outputs=96),
# op(ops.expanded_conv, stride=2, num_outputs=160),
# op(ops.expanded_conv, stride=1, num_outputs=160),
# op(ops.expanded_conv, stride=1, num_outputs=160),
# op(ops.expanded_conv, stride=1, num_outputs=320),
# op(slim.conv2d, stride=1, kernel_size=[1, 1], num_outputs=1280)
],
)
else:
V2_DEF = dict(
defaults={
# Note: these parameters of batch norm affect the architecture
# that's why they are here and not in training_scope.
(slim.batch_norm,): {'center': True, 'scale': True},
(slim.conv2d, slim.fully_connected, slim.separable_conv2d): {
'normalizer_fn': slim.batch_norm, 'activation_fn': tf.nn.relu6
},
(ops.expanded_conv,): {
'expansion_size': expand_input(6),
'split_expansion': 1,
'normalizer_fn': slim.batch_norm,
'residual': True
},
(slim.conv2d, slim.separable_conv2d): {'padding': 'SAME'}
},
spec=[
op(slim.conv2d, stride=2, num_outputs=32, kernel_size=[3, 3]),
op(ops.expanded_conv,
expansion_size=expand_input(1, divisible_by=1),
num_outputs=16),
op(ops.expanded_conv, stride=2, num_outputs=24),
op(ops.expanded_conv, stride=1, num_outputs=24),
# op(ops.expanded_conv, stride=2, num_outputs=32),
# op(ops.expanded_conv, stride=1, num_outputs=32),
# op(ops.expanded_conv, stride=1, num_outputs=32),
# op(ops.expanded_conv, stride=2, num_outputs=64),
# op(ops.expanded_conv, stride=1, num_outputs=64),
# op(ops.expanded_conv, stride=1, num_outputs=64),
# op(ops.expanded_conv, stride=1, num_outputs=64),
# op(ops.expanded_conv, stride=1, num_outputs=96),
# op(ops.expanded_conv, stride=1, num_outputs=96),
# op(ops.expanded_conv, stride=1, num_outputs=96),
# op(ops.expanded_conv, stride=2, num_outputs=160),
# op(ops.expanded_conv, stride=1, num_outputs=160),
# op(ops.expanded_conv, stride=1, num_outputs=160),
# op(ops.expanded_conv, stride=1, num_outputs=320),
# op(slim.conv2d, stride=1, kernel_size=[1, 1], num_outputs=1280)
],
)
if conv_defs is None:
conv_defs = V2_DEF
if 'multiplier' in kwargs:
raise ValueError('mobilenetv2 doesn\'t support generic '
'multiplier parameter use "depth_multiplier" instead.')
if finegrain_classification_mode:
conv_defs = copy.deepcopy(conv_defs)
if depth_multiplier < 1:
conv_defs['spec'][-1].params['num_outputs'] /= depth_multiplier
if activation_fn:
conv_defs = copy.deepcopy(conv_defs)
defaults = conv_defs['defaults']
conv_defaults = (
defaults[(slim.conv2d, slim.fully_connected, slim.separable_conv2d)])
conv_defaults['activation_fn'] = activation_fn
depth_args = {}
# NB: do not set depth_args unless they are provided to avoid overriding
# whatever default depth_multiplier might have thanks to arg_scope.
if min_depth is not None:
depth_args['min_depth'] = min_depth
if divisible_by is not None:
depth_args['divisible_by'] = divisible_by
with slim.arg_scope((lib.depth_multiplier,), **depth_args):
return lib.mobilenet(
input_tensor,
num_classes=num_classes,
conv_defs=conv_defs,
scope=scope,
multiplier=depth_multiplier,
**kwargs)
mobilenet.default_image_size = 224
def wrapped_partial(func, *args, **kwargs):
partial_func = functools.partial(func, *args, **kwargs)
functools.update_wrapper(partial_func, func)
return partial_func
# Wrappers for mobilenet v2 with depth-multipliers. Be noticed that
# 'finegrain_classification_mode' is set to True, which means the embedding
# layer will not be shrinked when given a depth-multiplier < 1.0.
mobilenet_v2_140 = wrapped_partial(mobilenet, depth_multiplier=1.4)
mobilenet_v2_050 = wrapped_partial(mobilenet, depth_multiplier=0.50,
finegrain_classification_mode=True)
mobilenet_v2_035 = wrapped_partial(mobilenet, depth_multiplier=0.35,
finegrain_classification_mode=True)
@slim.add_arg_scope
def mobilenet_base(input_tensor, depth_multiplier=1.0, **kwargs):
"""Creates base of the mobilenet (no pooling and no logits) ."""
return mobilenet(input_tensor,
depth_multiplier=depth_multiplier,
base_only=True, **kwargs)
def training_scope(**kwargs):
"""Defines MobilenetV2 training scope.
Usage:
with tf.contrib.slim.arg_scope(mobilenet_v2.training_scope()):
logits, endpoints = mobilenet_v2.mobilenet(input_tensor)
with slim.
Args:
**kwargs: Passed to mobilenet.training_scope. The following parameters
are supported:
weight_decay- The weight decay to use for regularizing the model.
stddev- Standard deviation for initialization, if negative uses xavier.
dropout_keep_prob- dropout keep probability
bn_decay- decay for the batch norm moving averages.
Returns:
An `arg_scope` to use for the mobilenet v2 model.
"""
return lib.training_scope(**kwargs)
__all__ = ['training_scope', 'mobilenet_base', 'mobilenet', 'V2_DEF']