model_builder.py 17.2 KB
Newer Older
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
# Copyright 2017 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.
# ==============================================================================

"""A function to build a DetectionModel from configuration."""
from object_detection.builders import anchor_generator_builder
from object_detection.builders import box_coder_builder
from object_detection.builders import box_predictor_builder
from object_detection.builders import hyperparams_builder
from object_detection.builders import image_resizer_builder
from object_detection.builders import losses_builder
from object_detection.builders import matcher_builder
from object_detection.builders import post_processing_builder
from object_detection.builders import region_similarity_calculator_builder as sim_calc
from object_detection.core import box_predictor
from object_detection.meta_architectures import faster_rcnn_meta_arch
from object_detection.meta_architectures import rfcn_meta_arch
from object_detection.meta_architectures import ssd_meta_arch
from object_detection.models import faster_rcnn_inception_resnet_v2_feature_extractor as frcnn_inc_res
31
32
from object_detection.models import faster_rcnn_inception_v2_feature_extractor as frcnn_inc_v2
from object_detection.models import faster_rcnn_nas_feature_extractor as frcnn_nas
33
from object_detection.models import faster_rcnn_pnas_feature_extractor as frcnn_pnas
34
from object_detection.models import faster_rcnn_resnet_v1_feature_extractor as frcnn_resnet_v1
35
from object_detection.models import ssd_resnet_v1_fpn_feature_extractor as ssd_resnet_v1_fpn
36
from object_detection.models.embedded_ssd_mobilenet_v1_feature_extractor import EmbeddedSSDMobileNetV1FeatureExtractor
37
from object_detection.models.ssd_inception_v2_feature_extractor import SSDInceptionV2FeatureExtractor
38
from object_detection.models.ssd_inception_v3_feature_extractor import SSDInceptionV3FeatureExtractor
39
from object_detection.models.ssd_mobilenet_v1_feature_extractor import SSDMobileNetV1FeatureExtractor
40
from object_detection.models.ssd_mobilenet_v2_feature_extractor import SSDMobileNetV2FeatureExtractor
41
42
43
44
45
from object_detection.protos import model_pb2

# A map of names to SSD feature extractors.
SSD_FEATURE_EXTRACTOR_CLASS_MAP = {
    'ssd_inception_v2': SSDInceptionV2FeatureExtractor,
46
    'ssd_inception_v3': SSDInceptionV3FeatureExtractor,
47
    'ssd_mobilenet_v1': SSDMobileNetV1FeatureExtractor,
48
    'ssd_mobilenet_v2': SSDMobileNetV2FeatureExtractor,
49
50
51
    'ssd_resnet50_v1_fpn': ssd_resnet_v1_fpn.SSDResnet50V1FpnFeatureExtractor,
    'ssd_resnet101_v1_fpn': ssd_resnet_v1_fpn.SSDResnet101V1FpnFeatureExtractor,
    'ssd_resnet152_v1_fpn': ssd_resnet_v1_fpn.SSDResnet152V1FpnFeatureExtractor,
52
    'embedded_ssd_mobilenet_v1': EmbeddedSSDMobileNetV1FeatureExtractor,
53
54
55
56
}

# A map of names to Faster R-CNN feature extractors.
FASTER_RCNN_FEATURE_EXTRACTOR_CLASS_MAP = {
Vivek Rathod's avatar
Vivek Rathod committed
57
58
    'faster_rcnn_nas':
    frcnn_nas.FasterRCNNNASFeatureExtractor,
59
60
    'faster_rcnn_pnas':
    frcnn_pnas.FasterRCNNPNASFeatureExtractor,
61
62
63
64
    'faster_rcnn_inception_resnet_v2':
    frcnn_inc_res.FasterRCNNInceptionResnetV2FeatureExtractor,
    'faster_rcnn_inception_v2':
    frcnn_inc_v2.FasterRCNNInceptionV2FeatureExtractor,
65
66
67
68
69
70
71
72
73
    'faster_rcnn_resnet50':
    frcnn_resnet_v1.FasterRCNNResnet50FeatureExtractor,
    'faster_rcnn_resnet101':
    frcnn_resnet_v1.FasterRCNNResnet101FeatureExtractor,
    'faster_rcnn_resnet152':
    frcnn_resnet_v1.FasterRCNNResnet152FeatureExtractor,
}


74
75
def build(model_config, is_training, add_summaries=True,
          add_background_class=True):
76
77
78
79
80
81
  """Builds a DetectionModel based on the model config.

  Args:
    model_config: A model.proto object containing the config for the desired
      DetectionModel.
    is_training: True if this model is being built for training purposes.
82
    add_summaries: Whether to add tensorflow summaries in the model graph.
83
84
85
86
    add_background_class: Whether to add an implicit background class to one-hot
      encodings of groundtruth labels. Set to false if using groundtruth labels
      with an explicit background class or using multiclass scores instead of
      truth in the case of distillation. Ignored in the case of faster_rcnn.
87
88
89
90
91
92
93
94
95
96
  Returns:
    DetectionModel based on the config.

  Raises:
    ValueError: On invalid meta architecture or model.
  """
  if not isinstance(model_config, model_pb2.DetectionModel):
    raise ValueError('model_config not of type model_pb2.DetectionModel.')
  meta_architecture = model_config.WhichOneof('model')
  if meta_architecture == 'ssd':
97
98
    return _build_ssd_model(model_config.ssd, is_training, add_summaries,
                            add_background_class)
99
  if meta_architecture == 'faster_rcnn':
100
101
    return _build_faster_rcnn_model(model_config.faster_rcnn, is_training,
                                    add_summaries)
102
103
104
105
  raise ValueError('Unknown meta architecture: {}'.format(meta_architecture))


def _build_ssd_feature_extractor(feature_extractor_config, is_training,
106
                                 reuse_weights=None):
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
  """Builds a ssd_meta_arch.SSDFeatureExtractor based on config.

  Args:
    feature_extractor_config: A SSDFeatureExtractor proto config from ssd.proto.
    is_training: True if this feature extractor is being built for training.
    reuse_weights: if the feature extractor should reuse weights.

  Returns:
    ssd_meta_arch.SSDFeatureExtractor based on config.

  Raises:
    ValueError: On invalid feature extractor type.
  """
  feature_type = feature_extractor_config.type
  depth_multiplier = feature_extractor_config.depth_multiplier
  min_depth = feature_extractor_config.min_depth
123
  pad_to_multiple = feature_extractor_config.pad_to_multiple
124
  use_explicit_padding = feature_extractor_config.use_explicit_padding
125
  use_depthwise = feature_extractor_config.use_depthwise
126
127
  conv_hyperparams = hyperparams_builder.build(
      feature_extractor_config.conv_hyperparams, is_training)
128
129
  override_base_feature_extractor_hyperparams = (
      feature_extractor_config.override_base_feature_extractor_hyperparams)
130
131
132
133
134

  if feature_type not in SSD_FEATURE_EXTRACTOR_CLASS_MAP:
    raise ValueError('Unknown ssd feature_extractor: {}'.format(feature_type))

  feature_extractor_class = SSD_FEATURE_EXTRACTOR_CLASS_MAP[feature_type]
135
136
  return feature_extractor_class(
      is_training, depth_multiplier, min_depth, pad_to_multiple,
137
138
      conv_hyperparams, reuse_weights, use_explicit_padding, use_depthwise,
      override_base_feature_extractor_hyperparams)
139
140


141
142
def _build_ssd_model(ssd_config, is_training, add_summaries,
                     add_background_class=True):
143
144
145
146
147
148
  """Builds an SSD detection model based on the model config.

  Args:
    ssd_config: A ssd.proto object containing the config for the desired
      SSDMetaArch.
    is_training: True if this model is being built for training purposes.
149
    add_summaries: Whether to add tf summaries in the model.
150
151
152
153
    add_background_class: Whether to add an implicit background class to one-hot
      encodings of groundtruth labels. Set to false if using groundtruth labels
      with an explicit background class or using multiclass scores instead of
      truth in the case of distillation.
154
155
  Returns:
    SSDMetaArch based on the config.
156

157
158
159
160
161
162
163
  Raises:
    ValueError: If ssd_config.type is not recognized (i.e. not registered in
      model_class_map).
  """
  num_classes = ssd_config.num_classes

  # Feature extractor
164
  feature_extractor = _build_ssd_feature_extractor(
165
      feature_extractor_config=ssd_config.feature_extractor,
166
      is_training=is_training)
167
168
169
170
171

  box_coder = box_coder_builder.build(ssd_config.box_coder)
  matcher = matcher_builder.build(ssd_config.matcher)
  region_similarity_calculator = sim_calc.build(
      ssd_config.similarity_calculator)
172
  encode_background_as_zeros = ssd_config.encode_background_as_zeros
173
  negative_class_weight = ssd_config.negative_class_weight
174
175
176
177
178
179
180
181
182
  ssd_box_predictor = box_predictor_builder.build(hyperparams_builder.build,
                                                  ssd_config.box_predictor,
                                                  is_training, num_classes)
  anchor_generator = anchor_generator_builder.build(
      ssd_config.anchor_generator)
  image_resizer_fn = image_resizer_builder.build(ssd_config.image_resizer)
  non_max_suppression_fn, score_conversion_fn = post_processing_builder.build(
      ssd_config.post_processing)
  (classification_loss, localization_loss, classification_weight,
183
184
   localization_weight, hard_example_miner,
   random_example_sampler) = losses_builder.build(ssd_config.loss)
185
  normalize_loss_by_num_matches = ssd_config.normalize_loss_by_num_matches
186
  normalize_loc_loss_by_codesize = ssd_config.normalize_loc_loss_by_codesize
187
188
189
190
191
192
193
194
195

  return ssd_meta_arch.SSDMetaArch(
      is_training,
      anchor_generator,
      ssd_box_predictor,
      box_coder,
      feature_extractor,
      matcher,
      region_similarity_calculator,
196
      encode_background_as_zeros,
197
      negative_class_weight,
198
199
200
201
202
203
204
205
      image_resizer_fn,
      non_max_suppression_fn,
      score_conversion_fn,
      classification_loss,
      localization_loss,
      classification_weight,
      localization_weight,
      normalize_loss_by_num_matches,
206
      hard_example_miner,
207
      add_summaries=add_summaries,
208
209
      normalize_loc_loss_by_codesize=normalize_loc_loss_by_codesize,
      freeze_batchnorm=ssd_config.freeze_batchnorm,
210
      inplace_batchnorm_update=ssd_config.inplace_batchnorm_update,
211
212
      add_background_class=add_background_class,
      random_example_sampler=random_example_sampler)
213
214
215


def _build_faster_rcnn_feature_extractor(
216
217
    feature_extractor_config, is_training, reuse_weights=None,
    inplace_batchnorm_update=False):
218
219
220
221
222
223
224
  """Builds a faster_rcnn_meta_arch.FasterRCNNFeatureExtractor based on config.

  Args:
    feature_extractor_config: A FasterRcnnFeatureExtractor proto config from
      faster_rcnn.proto.
    is_training: True if this feature extractor is being built for training.
    reuse_weights: if the feature extractor should reuse weights.
225
226
227
228
229
    inplace_batchnorm_update: Whether to update batch_norm inplace during
      training. This is required for batch norm to work correctly on TPUs. When
      this is false, user must add a control dependency on
      tf.GraphKeys.UPDATE_OPS for train/loss op in order to update the batch
      norm moving average parameters.
230
231
232
233
234
235
236

  Returns:
    faster_rcnn_meta_arch.FasterRCNNFeatureExtractor based on config.

  Raises:
    ValueError: On invalid feature extractor type.
  """
237
238
  if inplace_batchnorm_update:
    raise ValueError('inplace batchnorm updates not supported.')
239
240
241
  feature_type = feature_extractor_config.type
  first_stage_features_stride = (
      feature_extractor_config.first_stage_features_stride)
242
  batch_norm_trainable = feature_extractor_config.batch_norm_trainable
243
244
245
246
247
248
249

  if feature_type not in FASTER_RCNN_FEATURE_EXTRACTOR_CLASS_MAP:
    raise ValueError('Unknown Faster R-CNN feature_extractor: {}'.format(
        feature_type))
  feature_extractor_class = FASTER_RCNN_FEATURE_EXTRACTOR_CLASS_MAP[
      feature_type]
  return feature_extractor_class(
250
251
      is_training, first_stage_features_stride,
      batch_norm_trainable, reuse_weights)
252
253


254
def _build_faster_rcnn_model(frcnn_config, is_training, add_summaries):
255
256
257
258
259
260
261
  """Builds a Faster R-CNN or R-FCN detection model based on the model config.

  Builds R-FCN model if the second_stage_box_predictor in the config is of type
  `rfcn_box_predictor` else builds a Faster R-CNN model.

  Args:
    frcnn_config: A faster_rcnn.proto object containing the config for the
262
      desired FasterRCNNMetaArch or RFCNMetaArch.
263
    is_training: True if this model is being built for training purposes.
264
    add_summaries: Whether to add tf summaries in the model.
265
266
267

  Returns:
    FasterRCNNMetaArch based on the config.
268

269
270
271
272
273
274
275
276
  Raises:
    ValueError: If frcnn_config.type is not recognized (i.e. not registered in
      model_class_map).
  """
  num_classes = frcnn_config.num_classes
  image_resizer_fn = image_resizer_builder.build(frcnn_config.image_resizer)

  feature_extractor = _build_faster_rcnn_feature_extractor(
277
278
      frcnn_config.feature_extractor, is_training,
      frcnn_config.inplace_batchnorm_update)
279

280
  number_of_stages = frcnn_config.number_of_stages
281
282
283
284
  first_stage_anchor_generator = anchor_generator_builder.build(
      frcnn_config.first_stage_anchor_generator)

  first_stage_atrous_rate = frcnn_config.first_stage_atrous_rate
285
  first_stage_box_predictor_arg_scope_fn = hyperparams_builder.build(
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
      frcnn_config.first_stage_box_predictor_conv_hyperparams, is_training)
  first_stage_box_predictor_kernel_size = (
      frcnn_config.first_stage_box_predictor_kernel_size)
  first_stage_box_predictor_depth = frcnn_config.first_stage_box_predictor_depth
  first_stage_minibatch_size = frcnn_config.first_stage_minibatch_size
  first_stage_positive_balance_fraction = (
      frcnn_config.first_stage_positive_balance_fraction)
  first_stage_nms_score_threshold = frcnn_config.first_stage_nms_score_threshold
  first_stage_nms_iou_threshold = frcnn_config.first_stage_nms_iou_threshold
  first_stage_max_proposals = frcnn_config.first_stage_max_proposals
  first_stage_loc_loss_weight = (
      frcnn_config.first_stage_localization_loss_weight)
  first_stage_obj_loss_weight = frcnn_config.first_stage_objectness_loss_weight

  initial_crop_size = frcnn_config.initial_crop_size
  maxpool_kernel_size = frcnn_config.maxpool_kernel_size
  maxpool_stride = frcnn_config.maxpool_stride

  second_stage_box_predictor = box_predictor_builder.build(
      hyperparams_builder.build,
      frcnn_config.second_stage_box_predictor,
      is_training=is_training,
      num_classes=num_classes)
  second_stage_batch_size = frcnn_config.second_stage_batch_size
  second_stage_balance_fraction = frcnn_config.second_stage_balance_fraction
  (second_stage_non_max_suppression_fn, second_stage_score_conversion_fn
  ) = post_processing_builder.build(frcnn_config.second_stage_post_processing)
  second_stage_localization_loss_weight = (
      frcnn_config.second_stage_localization_loss_weight)
315
316
317
  second_stage_classification_loss = (
      losses_builder.build_faster_rcnn_classification_loss(
          frcnn_config.second_stage_classification_loss))
318
319
  second_stage_classification_loss_weight = (
      frcnn_config.second_stage_classification_loss_weight)
320
321
  second_stage_mask_prediction_loss_weight = (
      frcnn_config.second_stage_mask_prediction_loss_weight)
322
323
324
325
326
327
328
329
330
331
332
333
334

  hard_example_miner = None
  if frcnn_config.HasField('hard_example_miner'):
    hard_example_miner = losses_builder.build_hard_example_miner(
        frcnn_config.hard_example_miner,
        second_stage_classification_loss_weight,
        second_stage_localization_loss_weight)

  common_kwargs = {
      'is_training': is_training,
      'num_classes': num_classes,
      'image_resizer_fn': image_resizer_fn,
      'feature_extractor': feature_extractor,
335
      'number_of_stages': number_of_stages,
336
337
      'first_stage_anchor_generator': first_stage_anchor_generator,
      'first_stage_atrous_rate': first_stage_atrous_rate,
338
339
      'first_stage_box_predictor_arg_scope_fn':
      first_stage_box_predictor_arg_scope_fn,
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
      'first_stage_box_predictor_kernel_size':
      first_stage_box_predictor_kernel_size,
      'first_stage_box_predictor_depth': first_stage_box_predictor_depth,
      'first_stage_minibatch_size': first_stage_minibatch_size,
      'first_stage_positive_balance_fraction':
      first_stage_positive_balance_fraction,
      'first_stage_nms_score_threshold': first_stage_nms_score_threshold,
      'first_stage_nms_iou_threshold': first_stage_nms_iou_threshold,
      'first_stage_max_proposals': first_stage_max_proposals,
      'first_stage_localization_loss_weight': first_stage_loc_loss_weight,
      'first_stage_objectness_loss_weight': first_stage_obj_loss_weight,
      'second_stage_batch_size': second_stage_batch_size,
      'second_stage_balance_fraction': second_stage_balance_fraction,
      'second_stage_non_max_suppression_fn':
      second_stage_non_max_suppression_fn,
      'second_stage_score_conversion_fn': second_stage_score_conversion_fn,
      'second_stage_localization_loss_weight':
      second_stage_localization_loss_weight,
358
359
      'second_stage_classification_loss':
      second_stage_classification_loss,
360
361
      'second_stage_classification_loss_weight':
      second_stage_classification_loss_weight,
362
363
      'hard_example_miner': hard_example_miner,
      'add_summaries': add_summaries}
364
365
366
367
368
369
370
371
372
373
374

  if isinstance(second_stage_box_predictor, box_predictor.RfcnBoxPredictor):
    return rfcn_meta_arch.RFCNMetaArch(
        second_stage_rfcn_box_predictor=second_stage_box_predictor,
        **common_kwargs)
  else:
    return faster_rcnn_meta_arch.FasterRCNNMetaArch(
        initial_crop_size=initial_crop_size,
        maxpool_kernel_size=maxpool_kernel_size,
        maxpool_stride=maxpool_stride,
        second_stage_mask_rcnn_box_predictor=second_stage_box_predictor,
375
376
        second_stage_mask_prediction_loss_weight=(
            second_stage_mask_prediction_loss_weight),
377
        **common_kwargs)