model_lib.py 46.4 KB
Newer Older
1
2
3
4
5
6
7
8
9
10
11
12
13
14
# 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.
# ==============================================================================
15
r"""Constructs model, inputs, and training environment."""
16
17
18
19
20

from __future__ import absolute_import
from __future__ import division
from __future__ import print_function

21
import copy
22
import functools
23
import os
24

25
import tensorflow.compat.v1 as tf
26
import tensorflow.compat.v2 as tf2
27
28
import tf_slim as slim

29
from object_detection import eval_util
30
from object_detection import exporter as exporter_lib
31
from object_detection import inputs
32
from object_detection.builders import graph_rewriter_builder
33
34
35
36
37
from object_detection.builders import model_builder
from object_detection.builders import optimizer_builder
from object_detection.core import standard_fields as fields
from object_detection.utils import config_util
from object_detection.utils import label_map_util
38
from object_detection.utils import ops
39
40
41
42
from object_detection.utils import shape_utils
from object_detection.utils import variables_helper
from object_detection.utils import visualization_utils as vis_utils

43
44
45
46
47
48
49
50
# pylint: disable=g-import-not-at-top
try:
  from tensorflow.contrib import learn as contrib_learn
except ImportError:
  # TF 2.0 doesn't ship with contrib.
  pass
# pylint: enable=g-import-not-at-top

51
52
53
54
55
56
57
58
# A map of names to methods that help build the model.
MODEL_BUILD_UTIL_MAP = {
    'get_configs_from_pipeline_file':
        config_util.get_configs_from_pipeline_file,
    'create_pipeline_proto_from_configs':
        config_util.create_pipeline_proto_from_configs,
    'merge_external_params_with_configs':
        config_util.merge_external_params_with_configs,
59
60
61
62
63
64
    'create_train_input_fn':
        inputs.create_train_input_fn,
    'create_eval_input_fn':
        inputs.create_eval_input_fn,
    'create_predict_input_fn':
        inputs.create_predict_input_fn,
65
    'detection_model_fn_base': model_builder.build,
66
67
68
}


69
70
def _prepare_groundtruth_for_eval(detection_model, class_agnostic,
                                  max_number_of_boxes):
71
  """Extracts groundtruth data from detection_model and prepares it for eval.
72
73
74
75

  Args:
    detection_model: A `DetectionModel` object.
    class_agnostic: Whether the detections are class_agnostic.
76
    max_number_of_boxes: Max number of groundtruth boxes.
77
78
79
80

  Returns:
    A tuple of:
    groundtruth: Dictionary with the following fields:
81
82
83
84
85
      'groundtruth_boxes': [batch_size, num_boxes, 4] float32 tensor of boxes,
        in normalized coordinates.
      'groundtruth_classes': [batch_size, num_boxes] int64 tensor of 1-indexed
        classes.
      'groundtruth_masks': 4D float32 tensor of instance masks (if provided in
86
        groundtruth)
87
88
      'groundtruth_is_crowd': [batch_size, num_boxes] bool tensor indicating
        is_crowd annotations (if provided in groundtruth).
89
90
91
      'groundtruth_area': [batch_size, num_boxes] float32 tensor indicating
        the area (in the original absolute coordinates) of annotations (if
        provided in groundtruth).
92
93
      'num_groundtruth_boxes': [batch_size] tensor containing the maximum number
        of groundtruth boxes per image..
94
95
      'groundtruth_keypoints': [batch_size, num_boxes, num_keypoints, 2] float32
        tensor of keypoints (if provided in groundtruth).
96
97
98
99
100
101
102
103
104
      'groundtruth_dp_num_points_list': [batch_size, num_boxes] int32 tensor
        with the number of DensePose points for each instance (if provided in
        groundtruth).
      'groundtruth_dp_part_ids_list': [batch_size, num_boxes,
        max_sampled_points] int32 tensor with the part ids for each DensePose
        sampled point (if provided in groundtruth).
      'groundtruth_dp_surface_coords_list': [batch_size, num_boxes,
        max_sampled_points, 4] containing the DensePose surface coordinates for
        each sampled point (if provided in groundtruth).
105
106
      'groundtruth_track_ids_list': [batch_size, num_boxes] int32 tensor
        with track ID for each instance (if provided in groundtruth).
107
108
109
110
      'groundtruth_group_of': [batch_size, num_boxes] bool tensor indicating
        group_of annotations (if provided in groundtruth).
      'groundtruth_labeled_classes': [batch_size, num_classes] int64
        tensor of 1-indexed classes.
111
112
113
114
115
116
      'groundtruth_verified_neg_classes': [batch_size, num_classes] float32
        K-hot representation of 1-indexed classes which were verified as not
        present in the image.
      'groundtruth_not_exhaustive_classes': [batch_size, num_classes] K-hot
        representation of 1-indexed classes which don't have all of their
        instances marked exhaustively.
117
118
119
    class_agnostic: Boolean indicating whether detections are class agnostic.
  """
  input_data_fields = fields.InputDataFields()
120
121
122
  groundtruth_boxes = tf.stack(
      detection_model.groundtruth_lists(fields.BoxListFields.boxes))
  groundtruth_boxes_shape = tf.shape(groundtruth_boxes)
123
124
125
  # For class-agnostic models, groundtruth one-hot encodings collapse to all
  # ones.
  if class_agnostic:
126
127
    groundtruth_classes_one_hot = tf.ones(
        [groundtruth_boxes_shape[0], groundtruth_boxes_shape[1], 1])
128
  else:
129
130
    groundtruth_classes_one_hot = tf.stack(
        detection_model.groundtruth_lists(fields.BoxListFields.classes))
131
132
  label_id_offset = 1  # Applying label id offset (b/63711816)
  groundtruth_classes = (
133
      tf.argmax(groundtruth_classes_one_hot, axis=2) + label_id_offset)
134
135
136
137
  groundtruth = {
      input_data_fields.groundtruth_boxes: groundtruth_boxes,
      input_data_fields.groundtruth_classes: groundtruth_classes
  }
138

139
  if detection_model.groundtruth_has_field(fields.BoxListFields.masks):
140
141
142
    groundtruth[input_data_fields.groundtruth_instance_masks] = tf.stack(
        detection_model.groundtruth_lists(fields.BoxListFields.masks))

143
  if detection_model.groundtruth_has_field(fields.BoxListFields.is_crowd):
144
145
146
    groundtruth[input_data_fields.groundtruth_is_crowd] = tf.stack(
        detection_model.groundtruth_lists(fields.BoxListFields.is_crowd))

147
148
149
150
151
152
153
154
155
156
157
158
159
160
  if detection_model.groundtruth_has_field(input_data_fields.groundtruth_area):
    groundtruth[input_data_fields.groundtruth_area] = tf.stack(
        detection_model.groundtruth_lists(input_data_fields.groundtruth_area))

  if detection_model.groundtruth_has_field(fields.BoxListFields.keypoints):
    groundtruth[input_data_fields.groundtruth_keypoints] = tf.stack(
        detection_model.groundtruth_lists(fields.BoxListFields.keypoints))

  if detection_model.groundtruth_has_field(
      fields.BoxListFields.keypoint_visibilities):
    groundtruth[input_data_fields.groundtruth_keypoint_visibilities] = tf.stack(
        detection_model.groundtruth_lists(
            fields.BoxListFields.keypoint_visibilities))

161
162
163
164
  if detection_model.groundtruth_has_field(fields.BoxListFields.group_of):
    groundtruth[input_data_fields.groundtruth_group_of] = tf.stack(
        detection_model.groundtruth_lists(fields.BoxListFields.group_of))

165
  label_id_offset_paddings = tf.constant([[0, 0], [1, 0]])
166
  if detection_model.groundtruth_has_field(
167
      input_data_fields.groundtruth_verified_neg_classes):
168
169
170
171
    groundtruth[input_data_fields.groundtruth_verified_neg_classes] = tf.pad(
        tf.stack(detection_model.groundtruth_lists(
            input_data_fields.groundtruth_verified_neg_classes)),
        label_id_offset_paddings)
172
173
174
175

  if detection_model.groundtruth_has_field(
      input_data_fields.groundtruth_not_exhaustive_classes):
    groundtruth[
176
177
178
179
        input_data_fields.groundtruth_not_exhaustive_classes] = tf.pad(
            tf.stack(detection_model.groundtruth_lists(
                input_data_fields.groundtruth_not_exhaustive_classes)),
            label_id_offset_paddings)
180

181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
  if detection_model.groundtruth_has_field(
      fields.BoxListFields.densepose_num_points):
    groundtruth[input_data_fields.groundtruth_dp_num_points] = tf.stack(
        detection_model.groundtruth_lists(
            fields.BoxListFields.densepose_num_points))
  if detection_model.groundtruth_has_field(
      fields.BoxListFields.densepose_part_ids):
    groundtruth[input_data_fields.groundtruth_dp_part_ids] = tf.stack(
        detection_model.groundtruth_lists(
            fields.BoxListFields.densepose_part_ids))
  if detection_model.groundtruth_has_field(
      fields.BoxListFields.densepose_surface_coords):
    groundtruth[input_data_fields.groundtruth_dp_surface_coords] = tf.stack(
        detection_model.groundtruth_lists(
            fields.BoxListFields.densepose_surface_coords))
196
197
198
199
200

  if detection_model.groundtruth_has_field(fields.BoxListFields.track_ids):
    groundtruth[input_data_fields.groundtruth_track_ids] = tf.stack(
        detection_model.groundtruth_lists(fields.BoxListFields.track_ids))

201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
  if detection_model.groundtruth_has_field(
      input_data_fields.groundtruth_labeled_classes):
    labeled_classes_list = detection_model.groundtruth_lists(
        input_data_fields.groundtruth_labeled_classes)
    labeled_classes = [
        tf.where(x)[:, 0] + label_id_offset for x in labeled_classes_list
    ]
    if len(labeled_classes) > 1:
      num_classes = labeled_classes_list[0].shape[0]
      padded_labeled_classes = []
      for x in labeled_classes:
        padding = num_classes - tf.shape(x)[0]
        padded_labeled_classes.append(tf.pad(x, [[0, padding]]))
      groundtruth[input_data_fields.groundtruth_labeled_classes] = tf.stack(
          padded_labeled_classes)
    else:
      groundtruth[input_data_fields.groundtruth_labeled_classes] = tf.stack(
          labeled_classes)

220
221
  groundtruth[input_data_fields.num_groundtruth_boxes] = (
      tf.tile([max_number_of_boxes], multiples=[groundtruth_boxes_shape[0]]))
222
223
224
225
226
227
228
  return groundtruth


def unstack_batch(tensor_dict, unpad_groundtruth_tensors=True):
  """Unstacks all tensors in `tensor_dict` along 0th dimension.

  Unstacks tensor from the tensor dict along 0th dimension and returns a
229
  tensor_dict containing values that are lists of unstacked, unpadded tensors.
230
231
232
233
234
235

  Tensors in the `tensor_dict` are expected to be of one of the three shapes:
  1. [batch_size]
  2. [batch_size, height, width, channels]
  3. [batch_size, num_boxes, d1, d2, ... dn]

236
237
  When unpad_groundtruth_tensors is set to true, unstacked tensors of form 3
  above are sliced along the `num_boxes` dimension using the value in tensor
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
  field.InputDataFields.num_groundtruth_boxes.

  Note that this function has a static list of input data fields and has to be
  kept in sync with the InputDataFields defined in core/standard_fields.py

  Args:
    tensor_dict: A dictionary of batched groundtruth tensors.
    unpad_groundtruth_tensors: Whether to remove padding along `num_boxes`
      dimension of the groundtruth tensors.

  Returns:
    A dictionary where the keys are from fields.InputDataFields and values are
    a list of unstacked (optionally unpadded) tensors.

  Raises:
    ValueError: If unpad_tensors is True and `tensor_dict` does not contain
      `num_groundtruth_boxes` tensor.
  """
256
257
258
  unbatched_tensor_dict = {
      key: tf.unstack(tensor) for key, tensor in tensor_dict.items()
  }
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
  if unpad_groundtruth_tensors:
    if (fields.InputDataFields.num_groundtruth_boxes not in
        unbatched_tensor_dict):
      raise ValueError('`num_groundtruth_boxes` not found in tensor_dict. '
                       'Keys available: {}'.format(
                           unbatched_tensor_dict.keys()))
    unbatched_unpadded_tensor_dict = {}
    unpad_keys = set([
        # List of input data fields that are padded along the num_boxes
        # dimension. This list has to be kept in sync with InputDataFields in
        # standard_fields.py.
        fields.InputDataFields.groundtruth_instance_masks,
        fields.InputDataFields.groundtruth_classes,
        fields.InputDataFields.groundtruth_boxes,
        fields.InputDataFields.groundtruth_keypoints,
274
        fields.InputDataFields.groundtruth_keypoint_visibilities,
275
276
277
        fields.InputDataFields.groundtruth_dp_num_points,
        fields.InputDataFields.groundtruth_dp_part_ids,
        fields.InputDataFields.groundtruth_dp_surface_coords,
278
        fields.InputDataFields.groundtruth_track_ids,
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
        fields.InputDataFields.groundtruth_group_of,
        fields.InputDataFields.groundtruth_difficult,
        fields.InputDataFields.groundtruth_is_crowd,
        fields.InputDataFields.groundtruth_area,
        fields.InputDataFields.groundtruth_weights
    ]).intersection(set(unbatched_tensor_dict.keys()))

    for key in unpad_keys:
      unpadded_tensor_list = []
      for num_gt, padded_tensor in zip(
          unbatched_tensor_dict[fields.InputDataFields.num_groundtruth_boxes],
          unbatched_tensor_dict[key]):
        tensor_shape = shape_utils.combined_static_and_dynamic_shape(
            padded_tensor)
        slice_begin = tf.zeros([len(tensor_shape)], dtype=tf.int32)
        slice_size = tf.stack(
            [num_gt] + [-1 if dim is None else dim for dim in tensor_shape[1:]])
        unpadded_tensor = tf.slice(padded_tensor, slice_begin, slice_size)
        unpadded_tensor_list.append(unpadded_tensor)
      unbatched_unpadded_tensor_dict[key] = unpadded_tensor_list
299

300
301
302
303
304
    unbatched_tensor_dict.update(unbatched_unpadded_tensor_dict)

  return unbatched_tensor_dict


pkulzc's avatar
pkulzc committed
305
def provide_groundtruth(model, labels):
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
  """Provides the labels to a model as groundtruth.

  This helper function extracts the corresponding boxes, classes,
  keypoints, weights, masks, etc. from the labels, and provides it
  as groundtruth to the models.

  Args:
    model: The detection model to provide groundtruth to.
    labels: The labels for the training or evaluation inputs.
  """
  gt_boxes_list = labels[fields.InputDataFields.groundtruth_boxes]
  gt_classes_list = labels[fields.InputDataFields.groundtruth_classes]
  gt_masks_list = None
  if fields.InputDataFields.groundtruth_instance_masks in labels:
    gt_masks_list = labels[
        fields.InputDataFields.groundtruth_instance_masks]
  gt_keypoints_list = None
  if fields.InputDataFields.groundtruth_keypoints in labels:
    gt_keypoints_list = labels[fields.InputDataFields.groundtruth_keypoints]
325
326
327
328
  gt_keypoint_visibilities_list = None
  if fields.InputDataFields.groundtruth_keypoint_visibilities in labels:
    gt_keypoint_visibilities_list = labels[
        fields.InputDataFields.groundtruth_keypoint_visibilities]
329
330
331
332
333
334
335
336
337
338
339
340
  gt_dp_num_points_list = None
  if fields.InputDataFields.groundtruth_dp_num_points in labels:
    gt_dp_num_points_list = labels[
        fields.InputDataFields.groundtruth_dp_num_points]
  gt_dp_part_ids_list = None
  if fields.InputDataFields.groundtruth_dp_part_ids in labels:
    gt_dp_part_ids_list = labels[
        fields.InputDataFields.groundtruth_dp_part_ids]
  gt_dp_surface_coords_list = None
  if fields.InputDataFields.groundtruth_dp_surface_coords in labels:
    gt_dp_surface_coords_list = labels[
        fields.InputDataFields.groundtruth_dp_surface_coords]
341
342
343
344
  gt_track_ids_list = None
  if fields.InputDataFields.groundtruth_track_ids in labels:
    gt_track_ids_list = labels[
        fields.InputDataFields.groundtruth_track_ids]
345
346
347
348
349
350
351
352
353
354
  gt_weights_list = None
  if fields.InputDataFields.groundtruth_weights in labels:
    gt_weights_list = labels[fields.InputDataFields.groundtruth_weights]
  gt_confidences_list = None
  if fields.InputDataFields.groundtruth_confidences in labels:
    gt_confidences_list = labels[
        fields.InputDataFields.groundtruth_confidences]
  gt_is_crowd_list = None
  if fields.InputDataFields.groundtruth_is_crowd in labels:
    gt_is_crowd_list = labels[fields.InputDataFields.groundtruth_is_crowd]
355
356
357
  gt_group_of_list = None
  if fields.InputDataFields.groundtruth_group_of in labels:
    gt_group_of_list = labels[fields.InputDataFields.groundtruth_group_of]
358
359
360
361
362
363
364
  gt_area_list = None
  if fields.InputDataFields.groundtruth_area in labels:
    gt_area_list = labels[fields.InputDataFields.groundtruth_area]
  gt_labeled_classes = None
  if fields.InputDataFields.groundtruth_labeled_classes in labels:
    gt_labeled_classes = labels[
        fields.InputDataFields.groundtruth_labeled_classes]
365
366
367
368
369
370
371
372
  gt_verified_neg_classes = None
  if fields.InputDataFields.groundtruth_verified_neg_classes in labels:
    gt_verified_neg_classes = labels[
        fields.InputDataFields.groundtruth_verified_neg_classes]
  gt_not_exhaustive_classes = None
  if fields.InputDataFields.groundtruth_not_exhaustive_classes in labels:
    gt_not_exhaustive_classes = labels[
        fields.InputDataFields.groundtruth_not_exhaustive_classes]
373
374
375
376
  model.provide_groundtruth(
      groundtruth_boxes_list=gt_boxes_list,
      groundtruth_classes_list=gt_classes_list,
      groundtruth_confidences_list=gt_confidences_list,
377
      groundtruth_labeled_classes=gt_labeled_classes,
378
379
      groundtruth_masks_list=gt_masks_list,
      groundtruth_keypoints_list=gt_keypoints_list,
380
      groundtruth_keypoint_visibilities_list=gt_keypoint_visibilities_list,
381
382
383
      groundtruth_dp_num_points_list=gt_dp_num_points_list,
      groundtruth_dp_part_ids_list=gt_dp_part_ids_list,
      groundtruth_dp_surface_coords_list=gt_dp_surface_coords_list,
384
      groundtruth_weights_list=gt_weights_list,
385
      groundtruth_is_crowd_list=gt_is_crowd_list,
386
      groundtruth_group_of_list=gt_group_of_list,
387
      groundtruth_area_list=gt_area_list,
388
389
390
      groundtruth_track_ids_list=gt_track_ids_list,
      groundtruth_verified_neg_classes=gt_verified_neg_classes,
      groundtruth_not_exhaustive_classes=gt_not_exhaustive_classes)
391
392


393
def create_model_fn(detection_model_fn, configs, hparams=None, use_tpu=False,
394
                    postprocess_on_cpu=False):
395
396
397
398
399
400
401
402
  """Creates a model function for `Estimator`.

  Args:
    detection_model_fn: Function that returns a `DetectionModel` instance.
    configs: Dictionary of pipeline config objects.
    hparams: `HParams` object.
    use_tpu: Boolean indicating whether model should be constructed for
        use on TPU.
403
404
    postprocess_on_cpu: When use_tpu and postprocess_on_cpu is true, postprocess
        is scheduled on the host cpu.
405
406
407
408
409
410

  Returns:
    `model_fn` for `Estimator`.
  """
  train_config = configs['train_config']
  eval_input_config = configs['eval_input_config']
411
  eval_config = configs['eval_config']
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429

  def model_fn(features, labels, mode, params=None):
    """Constructs the object detection model.

    Args:
      features: Dictionary of feature tensors, returned from `input_fn`.
      labels: Dictionary of groundtruth tensors if mode is TRAIN or EVAL,
        otherwise None.
      mode: Mode key from tf.estimator.ModeKeys.
      params: Parameter dictionary passed from the estimator.

    Returns:
      An `EstimatorSpec` that encapsulates the model and its serving
        configurations.
    """
    params = params or {}
    total_loss, train_op, detections, export_outputs = None, None, None, None
    is_training = mode == tf.estimator.ModeKeys.TRAIN
430
431
432
433

    # Make sure to set the Keras learning phase. True during training,
    # False for inference.
    tf.keras.backend.set_learning_phase(is_training)
434
435
436
437
438
    # Set policy for mixed-precision training with Keras-based models.
    if use_tpu and train_config.use_bfloat16:
      from tensorflow.python.keras.engine import base_layer_utils  # pylint: disable=g-import-not-at-top
      # Enable v2 behavior, as `mixed_bfloat16` is only supported in TF 2.0.
      base_layer_utils.enable_v2_dtype_behavior()
439
      tf2.keras.mixed_precision.experimental.set_policy(
440
          'mixed_bfloat16')
441
442
    detection_model = detection_model_fn(
        is_training=is_training, add_summaries=(not use_tpu))
443
444
445
446
447
448
449
    scaffold_fn = None

    if mode == tf.estimator.ModeKeys.TRAIN:
      labels = unstack_batch(
          labels,
          unpad_groundtruth_tensors=train_config.unpad_groundtruth_tensors)
    elif mode == tf.estimator.ModeKeys.EVAL:
450
451
452
453
454
      # For evaling on train data, it is necessary to check whether groundtruth
      # must be unpadded.
      boxes_shape = (
          labels[fields.InputDataFields.groundtruth_boxes].get_shape()
          .as_list())
455
      unpad_groundtruth_tensors = boxes_shape[1] is not None and not use_tpu
456
457
      labels = unstack_batch(
          labels, unpad_groundtruth_tensors=unpad_groundtruth_tensors)
458
459

    if mode in (tf.estimator.ModeKeys.TRAIN, tf.estimator.ModeKeys.EVAL):
pkulzc's avatar
pkulzc committed
460
      provide_groundtruth(detection_model, labels)
461
462

    preprocessed_images = features[fields.InputDataFields.image]
463
464
465

    side_inputs = detection_model.get_side_inputs(features)

466
    if use_tpu and train_config.use_bfloat16:
467
      with tf.tpu.bfloat16_scope():
468
469
        prediction_dict = detection_model.predict(
            preprocessed_images,
470
            features[fields.InputDataFields.true_image_shape], **side_inputs)
471
        prediction_dict = ops.bfloat16_to_float32_nested(prediction_dict)
472
473
474
    else:
      prediction_dict = detection_model.predict(
          preprocessed_images,
475
          features[fields.InputDataFields.true_image_shape], **side_inputs)
476
477
478
479

    def postprocess_wrapper(args):
      return detection_model.postprocess(args[0], args[1])

480
    if mode in (tf.estimator.ModeKeys.EVAL, tf.estimator.ModeKeys.PREDICT):
481
      if use_tpu and postprocess_on_cpu:
482
        detections = tf.tpu.outside_compilation(
483
484
485
486
487
488
489
            postprocess_wrapper,
            (prediction_dict,
             features[fields.InputDataFields.true_image_shape]))
      else:
        detections = postprocess_wrapper((
            prediction_dict,
            features[fields.InputDataFields.true_image_shape]))
490
491

    if mode == tf.estimator.ModeKeys.TRAIN:
492
493
      load_pretrained = hparams.load_pretrained if hparams else False
      if train_config.fine_tune_checkpoint and load_pretrained:
494
495
496
497
498
499
500
501
        if not train_config.fine_tune_checkpoint_type:
          # train_config.from_detection_checkpoint field is deprecated. For
          # backward compatibility, set train_config.fine_tune_checkpoint_type
          # based on train_config.from_detection_checkpoint.
          if train_config.from_detection_checkpoint:
            train_config.fine_tune_checkpoint_type = 'detection'
          else:
            train_config.fine_tune_checkpoint_type = 'classification'
502
        asg_map = detection_model.restore_map(
503
            fine_tune_checkpoint_type=train_config.fine_tune_checkpoint_type,
504
505
506
507
            load_all_detection_checkpoint_vars=(
                train_config.load_all_detection_checkpoint_vars))
        available_var_map = (
            variables_helper.get_variables_available_in_checkpoint(
508
509
                asg_map,
                train_config.fine_tune_checkpoint,
510
511
                include_global_step=False))
        if use_tpu:
512

513
514
515
516
          def tpu_scaffold():
            tf.train.init_from_checkpoint(train_config.fine_tune_checkpoint,
                                          available_var_map)
            return tf.train.Scaffold()
517

518
519
520
521
522
523
          scaffold_fn = tpu_scaffold
        else:
          tf.train.init_from_checkpoint(train_config.fine_tune_checkpoint,
                                        available_var_map)

    if mode in (tf.estimator.ModeKeys.TRAIN, tf.estimator.ModeKeys.EVAL):
524
525
526
527
528
529
530
531
532
533
534
535
536
537
538
539
540
541
542
543
      if (mode == tf.estimator.ModeKeys.EVAL and
          eval_config.use_dummy_loss_in_eval):
        total_loss = tf.constant(1.0)
        losses_dict = {'Loss/total_loss': total_loss}
      else:
        losses_dict = detection_model.loss(
            prediction_dict, features[fields.InputDataFields.true_image_shape])
        losses = [loss_tensor for loss_tensor in losses_dict.values()]
        if train_config.add_regularization_loss:
          regularization_losses = detection_model.regularization_losses()
          if use_tpu and train_config.use_bfloat16:
            regularization_losses = ops.bfloat16_to_float32_nested(
                regularization_losses)
          if regularization_losses:
            regularization_loss = tf.add_n(
                regularization_losses, name='regularization_loss')
            losses.append(regularization_loss)
            losses_dict['Loss/regularization_loss'] = regularization_loss
        total_loss = tf.add_n(losses, name='total_loss')
        losses_dict['Loss/total_loss'] = total_loss
544

545
546
547
548
549
      if 'graph_rewriter_config' in configs:
        graph_rewriter_fn = graph_rewriter_builder.build(
            configs['graph_rewriter_config'], is_training=is_training)
        graph_rewriter_fn()

550
551
      # TODO(rathodv): Stop creating optimizer summary vars in EVAL mode once we
      # can write learning rate summaries on TPU without host calls.
552
553
554
555
      global_step = tf.train.get_or_create_global_step()
      training_optimizer, optimizer_summary_vars = optimizer_builder.build(
          train_config.optimizer)

556
    if mode == tf.estimator.ModeKeys.TRAIN:
557
      if use_tpu:
558
        training_optimizer = tf.tpu.CrossShardOptimizer(training_optimizer)
559
560
561

      # Optionally freeze some layers by setting their gradients to be zero.
      trainable_variables = None
562
563
564
565
566
567
      include_variables = (
          train_config.update_trainable_variables
          if train_config.update_trainable_variables else None)
      exclude_variables = (
          train_config.freeze_variables
          if train_config.freeze_variables else None)
568
      trainable_variables = slim.filter_variables(
569
570
571
          tf.trainable_variables(),
          include_patterns=include_variables,
          exclude_patterns=exclude_variables)
572
573
574
575
576
577
578
579
580

      clip_gradients_value = None
      if train_config.gradient_clipping_by_norm > 0:
        clip_gradients_value = train_config.gradient_clipping_by_norm

      if not use_tpu:
        for var in optimizer_summary_vars:
          tf.summary.scalar(var.op.name, var)
      summaries = [] if use_tpu else None
581
582
      if train_config.summarize_gradients:
        summaries = ['gradients', 'gradient_norm', 'global_gradient_norm']
583
      train_op = slim.optimizers.optimize_loss(
584
585
586
587
588
          loss=total_loss,
          global_step=global_step,
          learning_rate=None,
          clip_gradients=clip_gradients_value,
          optimizer=training_optimizer,
589
          update_ops=detection_model.updates(),
590
591
592
593
594
          variables=trainable_variables,
          summaries=summaries,
          name='')  # Preventing scope prefix on all variables.

    if mode == tf.estimator.ModeKeys.PREDICT:
595
      exported_output = exporter_lib.add_output_tensor_nodes(detections)
596
597
      export_outputs = {
          tf.saved_model.signature_constants.PREDICT_METHOD_NAME:
598
              tf.estimator.export.PredictOutput(exported_output)
599
600
601
      }

    eval_metric_ops = None
602
    scaffold = None
603
    if mode == tf.estimator.ModeKeys.EVAL:
604
605
      class_agnostic = (
          fields.DetectionResultFields.detection_classes not in detections)
606
607
608
      groundtruth = _prepare_groundtruth_for_eval(
          detection_model, class_agnostic,
          eval_input_config.max_number_of_boxes)
609
      use_original_images = fields.InputDataFields.original_image in features
pkulzc's avatar
pkulzc committed
610
      if use_original_images:
611
612
613
614
615
        eval_images = features[fields.InputDataFields.original_image]
        true_image_shapes = tf.slice(
            features[fields.InputDataFields.true_image_shape], [0, 0], [-1, 3])
        original_image_spatial_shapes = features[fields.InputDataFields
                                                 .original_image_spatial_shape]
pkulzc's avatar
pkulzc committed
616
617
      else:
        eval_images = features[fields.InputDataFields.image]
618
619
        true_image_shapes = None
        original_image_spatial_shapes = None
pkulzc's avatar
pkulzc committed
620

621
622
623
      eval_dict = eval_util.result_dict_for_batched_example(
          eval_images,
          features[inputs.HASH_KEY],
624
625
626
          detections,
          groundtruth,
          class_agnostic=class_agnostic,
627
628
629
          scale_to_absolute=True,
          original_image_spatial_shapes=original_image_spatial_shapes,
          true_image_shapes=true_image_shapes)
630

631
632
633
634
      if fields.InputDataFields.image_additional_channels in features:
        eval_dict[fields.InputDataFields.image_additional_channels] = features[
            fields.InputDataFields.image_additional_channels]

635
636
637
638
639
      if class_agnostic:
        category_index = label_map_util.create_class_agnostic_category_index()
      else:
        category_index = label_map_util.create_category_index_from_labelmap(
            eval_input_config.label_map_path)
640
      vis_metric_ops = None
641
      if not use_tpu and use_original_images:
642
643
644
        keypoint_edges = [
            (kp.start, kp.end) for kp in eval_config.keypoint_edge]

645
646
647
648
649
        eval_metric_op_vis = vis_utils.VisualizeSingleFrameDetections(
            category_index,
            max_examples_to_draw=eval_config.num_visualizations,
            max_boxes_to_draw=eval_config.max_num_boxes_to_visualize,
            min_score_thresh=eval_config.min_score_threshold,
650
651
            use_normalized_coordinates=False,
            keypoint_edges=keypoint_edges or None)
652
653
        vis_metric_ops = eval_metric_op_vis.get_estimator_eval_metric_ops(
            eval_dict)
654

655
656
      # Eval metrics on a single example.
      eval_metric_ops = eval_util.get_eval_metric_ops_for_evaluators(
DefineFC's avatar
DefineFC committed
657
          eval_config, list(category_index.values()), eval_dict)
658
659
660
661
      for loss_key, loss_tensor in iter(losses_dict.items()):
        eval_metric_ops[loss_key] = tf.metrics.mean(loss_tensor)
      for var in optimizer_summary_vars:
        eval_metric_ops[var.op.name] = (var, tf.no_op())
662
663
      if vis_metric_ops is not None:
        eval_metric_ops.update(vis_metric_ops)
664
      eval_metric_ops = {str(k): v for k, v in eval_metric_ops.items()}
665

666
667
668
669
670
671
672
673
674
675
      if eval_config.use_moving_averages:
        variable_averages = tf.train.ExponentialMovingAverage(0.0)
        variables_to_restore = variable_averages.variables_to_restore()
        keep_checkpoint_every_n_hours = (
            train_config.keep_checkpoint_every_n_hours)
        saver = tf.train.Saver(
            variables_to_restore,
            keep_checkpoint_every_n_hours=keep_checkpoint_every_n_hours)
        scaffold = tf.train.Scaffold(saver=saver)

676
677
    # EVAL executes on CPU, so use regular non-TPU EstimatorSpec.
    if use_tpu and mode != tf.estimator.ModeKeys.EVAL:
678
      return tf.estimator.tpu.TPUEstimatorSpec(
679
680
681
682
683
684
685
686
          mode=mode,
          scaffold_fn=scaffold_fn,
          predictions=detections,
          loss=total_loss,
          train_op=train_op,
          eval_metrics=eval_metric_ops,
          export_outputs=export_outputs)
    else:
687
688
689
690
691
692
693
694
695
      if scaffold is None:
        keep_checkpoint_every_n_hours = (
            train_config.keep_checkpoint_every_n_hours)
        saver = tf.train.Saver(
            sharded=True,
            keep_checkpoint_every_n_hours=keep_checkpoint_every_n_hours,
            save_relative_paths=True)
        tf.add_to_collection(tf.GraphKeys.SAVERS, saver)
        scaffold = tf.train.Scaffold(saver=saver)
696
697
698
699
700
701
      return tf.estimator.EstimatorSpec(
          mode=mode,
          predictions=detections,
          loss=total_loss,
          train_op=train_op,
          eval_metric_ops=eval_metric_ops,
702
703
          export_outputs=export_outputs,
          scaffold=scaffold)
704
705
706
707

  return model_fn


708
def create_estimator_and_inputs(run_config,
709
710
                                hparams=None,
                                pipeline_config_path=None,
711
                                config_override=None,
712
                                train_steps=None,
713
                                sample_1_of_n_eval_examples=1,
714
                                sample_1_of_n_eval_on_train_examples=1,
715
716
717
718
719
                                model_fn_creator=create_model_fn,
                                use_tpu_estimator=False,
                                use_tpu=False,
                                num_shards=1,
                                params=None,
720
                                override_eval_num_epochs=True,
721
                                save_final_config=False,
722
723
                                postprocess_on_cpu=False,
                                export_to_tpu=None,
724
725
                                **kwargs):
  """Creates `Estimator`, input functions, and steps.
726
727
728

  Args:
    run_config: A `RunConfig`.
729
    hparams: (optional) A `HParams`.
730
    pipeline_config_path: A path to a pipeline config file.
731
732
    config_override: A pipeline_pb2.TrainEvalPipelineConfig text proto to
      override the config from `pipeline_config_path`.
733
734
    train_steps: Number of training steps. If None, the number of training steps
      is set from the `TrainConfig` proto.
735
736
737
738
739
    sample_1_of_n_eval_examples: Integer representing how often an eval example
      should be sampled. If 1, will sample all examples.
    sample_1_of_n_eval_on_train_examples: Similar to
      `sample_1_of_n_eval_examples`, except controls the sampling of training
      data for evaluation.
740
741
742
743
744
745
746
747
748
749
    model_fn_creator: A function that creates a `model_fn` for `Estimator`.
      Follows the signature:

      * Args:
        * `detection_model_fn`: Function that returns `DetectionModel` instance.
        * `configs`: Dictionary of pipeline config objects.
        * `hparams`: `HParams` object.
      * Returns:
        `model_fn` for `Estimator`.

750
751
752
753
754
755
756
757
    use_tpu_estimator: Whether a `TPUEstimator` should be returned. If False,
      an `Estimator` will be returned.
    use_tpu: Boolean, whether training and evaluation should run on TPU. Only
      used if `use_tpu_estimator` is True.
    num_shards: Number of shards (TPU cores). Only used if `use_tpu_estimator`
      is True.
    params: Parameter dictionary passed from the estimator. Only used if
      `use_tpu_estimator` is True.
758
759
    override_eval_num_epochs: Whether to overwrite the number of epochs to 1 for
      eval_input.
760
761
    save_final_config: Whether to save final config (obtained after applying
      overrides) to `estimator.model_dir`.
762
763
764
765
766
    postprocess_on_cpu: When use_tpu and postprocess_on_cpu are true,
      postprocess is scheduled on the host cpu.
    export_to_tpu: When use_tpu and export_to_tpu are true,
      `export_savedmodel()` exports a metagraph for serving on TPU besides the
      one on CPU.
767
768
769
    **kwargs: Additional keyword arguments for configuration override.

  Returns:
770
771
772
    A dictionary with the following fields:
    'estimator': An `Estimator` or `TPUEstimator`.
    'train_input_fn': A training input function.
773
774
    'eval_input_fns': A list of all evaluation input functions.
    'eval_input_names': A list of names for each evaluation input.
775
    'eval_on_train_input_fn': An evaluation-on-train input function.
776
777
778
    'predict_input_fn': A prediction input function.
    'train_steps': Number of training steps. Either directly from input or from
      configuration.
779
  """
780
781
782
783
  get_configs_from_pipeline_file = MODEL_BUILD_UTIL_MAP[
      'get_configs_from_pipeline_file']
  merge_external_params_with_configs = MODEL_BUILD_UTIL_MAP[
      'merge_external_params_with_configs']
784
785
  create_pipeline_proto_from_configs = MODEL_BUILD_UTIL_MAP[
      'create_pipeline_proto_from_configs']
786
787
788
  create_train_input_fn = MODEL_BUILD_UTIL_MAP['create_train_input_fn']
  create_eval_input_fn = MODEL_BUILD_UTIL_MAP['create_eval_input_fn']
  create_predict_input_fn = MODEL_BUILD_UTIL_MAP['create_predict_input_fn']
789
  detection_model_fn_base = MODEL_BUILD_UTIL_MAP['detection_model_fn_base']
790

791
792
  configs = get_configs_from_pipeline_file(
      pipeline_config_path, config_override=config_override)
793
794
  kwargs.update({
      'train_steps': train_steps,
795
      'use_bfloat16': configs['train_config'].use_bfloat16 and use_tpu
796
  })
pkulzc's avatar
pkulzc committed
797
798
799
800
  if sample_1_of_n_eval_examples >= 1:
    kwargs.update({
        'sample_1_of_n_eval_examples': sample_1_of_n_eval_examples
    })
801
802
803
804
  if override_eval_num_epochs:
    kwargs.update({'eval_num_epochs': 1})
    tf.logging.warning(
        'Forced number of epochs for all eval validations to be 1.')
805
  configs = merge_external_params_with_configs(
806
      configs, hparams, kwargs_dict=kwargs)
807
808
809
810
  model_config = configs['model']
  train_config = configs['train_config']
  train_input_config = configs['train_input_config']
  eval_config = configs['eval_config']
811
812
813
814
815
816
817
818
819
820
821
  eval_input_configs = configs['eval_input_configs']
  eval_on_train_input_config = copy.deepcopy(train_input_config)
  eval_on_train_input_config.sample_1_of_n_examples = (
      sample_1_of_n_eval_on_train_examples)
  if override_eval_num_epochs and eval_on_train_input_config.num_epochs != 1:
    tf.logging.warning('Expected number of evaluation epochs is 1, but '
                       'instead encountered `eval_on_train_input_config'
                       '.num_epochs` = '
                       '{}. Overwriting `num_epochs` to 1.'.format(
                           eval_on_train_input_config.num_epochs))
    eval_on_train_input_config.num_epochs = 1
822

823
824
825
  # update train_steps from config but only when non-zero value is provided
  if train_steps is None and train_config.num_steps != 0:
    train_steps = train_config.num_steps
826
827

  detection_model_fn = functools.partial(
828
      detection_model_fn_base, model_config=model_config)
829

830
  # Create the input functions for TRAIN/EVAL/PREDICT.
831
  train_input_fn = create_train_input_fn(
832
833
834
      train_config=train_config,
      train_input_config=train_input_config,
      model_config=model_config)
835
836
837
838
839
840
841
842
843
  eval_input_fns = [
      create_eval_input_fn(
          eval_config=eval_config,
          eval_input_config=eval_input_config,
          model_config=model_config) for eval_input_config in eval_input_configs
  ]
  eval_input_names = [
      eval_input_config.name for eval_input_config in eval_input_configs
  ]
844
845
  eval_on_train_input_fn = create_eval_input_fn(
      eval_config=eval_config,
846
      eval_input_config=eval_on_train_input_config,
847
      model_config=model_config)
848
  predict_input_fn = create_predict_input_fn(
849
      model_config=model_config, predict_input_config=eval_input_configs[0])
850

851
  # Read export_to_tpu from hparams if not passed.
852
  if export_to_tpu is None and hparams is not None:
853
    export_to_tpu = hparams.get('export_to_tpu', False)
854
855
  tf.logging.info('create_estimator_and_inputs: use_tpu %s, export_to_tpu %s',
                  use_tpu, export_to_tpu)
856
857
  model_fn = model_fn_creator(detection_model_fn, configs, hparams, use_tpu,
                              postprocess_on_cpu)
858
  if use_tpu_estimator:
859
    estimator = tf.estimator.tpu.TPUEstimator(
860
861
862
863
864
865
        model_fn=model_fn,
        train_batch_size=train_config.batch_size,
        # For each core, only batch size 1 is supported for eval.
        eval_batch_size=num_shards * 1 if use_tpu else 1,
        use_tpu=use_tpu,
        config=run_config,
866
867
        export_to_tpu=export_to_tpu,
        eval_on_tpu=False,  # Eval runs on CPU, so disable eval on TPU
pkulzc's avatar
pkulzc committed
868
        params=params if params else {})
869
870
  else:
    estimator = tf.estimator.Estimator(model_fn=model_fn, config=run_config)
871

872
  # Write the as-run pipeline config to disk.
873
  if run_config.is_chief and save_final_config:
874
    pipeline_config_final = create_pipeline_proto_from_configs(configs)
875
    config_util.save_pipeline_config(pipeline_config_final, estimator.model_dir)
876

877
  return dict(
878
879
      estimator=estimator,
      train_input_fn=train_input_fn,
880
881
      eval_input_fns=eval_input_fns,
      eval_input_names=eval_input_names,
882
      eval_on_train_input_fn=eval_on_train_input_fn,
883
      predict_input_fn=predict_input_fn,
884
      train_steps=train_steps)
885
886
887


def create_train_and_eval_specs(train_input_fn,
888
                                eval_input_fns,
889
                                eval_on_train_input_fn,
890
891
892
893
                                predict_input_fn,
                                train_steps,
                                eval_on_train_data=False,
                                final_exporter_name='Servo',
894
                                eval_spec_names=None):
895
896
897
898
  """Creates a `TrainSpec` and `EvalSpec`s.

  Args:
    train_input_fn: Function that produces features and labels on train data.
899
900
    eval_input_fns: A list of functions that produce features and labels on eval
      data.
901
902
    eval_on_train_input_fn: Function that produces features and labels for
      evaluation on train data.
903
904
905
906
907
    predict_input_fn: Function that produces features for inference.
    train_steps: Number of training steps.
    eval_on_train_data: Whether to evaluate model on training data. Default is
      False.
    final_exporter_name: String name given to `FinalExporter`.
908
    eval_spec_names: A list of string names for each `EvalSpec`.
909
910

  Returns:
911
912
913
    Tuple of `TrainSpec` and list of `EvalSpecs`. If `eval_on_train_data` is
    True, the last `EvalSpec` in the list will correspond to training data. The
    rest EvalSpecs in the list are evaluation datas.
914
915
916
917
  """
  train_spec = tf.estimator.TrainSpec(
      input_fn=train_input_fn, max_steps=train_steps)

918
  if eval_spec_names is None:
919
    eval_spec_names = [str(i) for i in range(len(eval_input_fns))]
920
921

  eval_specs = []
922
923
924
925
926
927
928
929
  for index, (eval_spec_name, eval_input_fn) in enumerate(
      zip(eval_spec_names, eval_input_fns)):
    # Uses final_exporter_name as exporter_name for the first eval spec for
    # backward compatibility.
    if index == 0:
      exporter_name = final_exporter_name
    else:
      exporter_name = '{}_{}'.format(final_exporter_name, eval_spec_name)
930
931
932
933
934
935
936
937
    exporter = tf.estimator.FinalExporter(
        name=exporter_name, serving_input_receiver_fn=predict_input_fn)
    eval_specs.append(
        tf.estimator.EvalSpec(
            name=eval_spec_name,
            input_fn=eval_input_fn,
            steps=None,
            exporters=exporter))
938
939
940
941

  if eval_on_train_data:
    eval_specs.append(
        tf.estimator.EvalSpec(
942
            name='eval_on_train', input_fn=eval_on_train_input_fn, steps=None))
943
944

  return train_spec, eval_specs
945
946


947
948
949
950
951
952
953
954
955
956
957
958
959
960
961
962
963
964
965
966
967
968
969
970
971
972
973
974
975
976
977
978
979
980
981
982
983
984
985
986
987
988
def _evaluate_checkpoint(estimator,
                         input_fn,
                         checkpoint_path,
                         name,
                         max_retries=0):
  """Evaluates a checkpoint.

  Args:
    estimator: Estimator object to use for evaluation.
    input_fn: Input function to use for evaluation.
    checkpoint_path: Path of the checkpoint to evaluate.
    name: Namescope for eval summary.
    max_retries: Maximum number of times to retry the evaluation on encountering
      a tf.errors.InvalidArgumentError. If negative, will always retry the
      evaluation.

  Returns:
    Estimator evaluation results.
  """
  always_retry = True if max_retries < 0 else False
  retries = 0
  while always_retry or retries <= max_retries:
    try:
      return estimator.evaluate(
          input_fn=input_fn,
          steps=None,
          checkpoint_path=checkpoint_path,
          name=name)
    except tf.errors.InvalidArgumentError as e:
      if always_retry or retries < max_retries:
        tf.logging.info('Retrying checkpoint evaluation after exception: %s', e)
        retries += 1
      else:
        raise e


def continuous_eval(estimator,
                    model_dir,
                    input_fn,
                    train_steps,
                    name,
                    max_retries=0):
989
990
991
992
993
994
995
996
997
  """Perform continuous evaluation on checkpoints written to a model directory.

  Args:
    estimator: Estimator object to use for evaluation.
    model_dir: Model directory to read checkpoints for continuous evaluation.
    input_fn: Input function to use for evaluation.
    train_steps: Number of training steps. This is used to infer the last
      checkpoint and stop evaluation loop.
    name: Namescope for eval summary.
998
999
1000
    max_retries: Maximum number of times to retry the evaluation on encountering
      a tf.errors.InvalidArgumentError. If negative, will always retry the
      evaluation.
1001
  """
1002

1003
1004
1005
1006
  def terminate_eval():
    tf.logging.info('Terminating eval after 180 seconds of no checkpoints')
    return True

1007
  for ckpt in tf.train.checkpoints_iterator(
1008
1009
1010
1011
1012
      model_dir, min_interval_secs=180, timeout=None,
      timeout_fn=terminate_eval):

    tf.logging.info('Starting Evaluation.')
    try:
1013
1014
1015
1016
1017
1018
      eval_results = _evaluate_checkpoint(
          estimator=estimator,
          input_fn=input_fn,
          checkpoint_path=ckpt,
          name=name,
          max_retries=max_retries)
1019
1020
1021
1022
1023
1024
1025
1026
1027
1028
1029
1030
1031
1032
      tf.logging.info('Eval results: %s' % eval_results)

      # Terminate eval job when final checkpoint is reached
      current_step = int(os.path.basename(ckpt).split('-')[1])
      if current_step >= train_steps:
        tf.logging.info(
            'Evaluation finished after training step %d' % current_step)
        break

    except tf.errors.NotFoundError:
      tf.logging.info(
          'Checkpoint %s no longer exists, skipping checkpoint' % ckpt)


1033
1034
1035
1036
1037
1038
1039
1040
def populate_experiment(run_config,
                        hparams,
                        pipeline_config_path,
                        train_steps=None,
                        eval_steps=None,
                        model_fn_creator=create_model_fn,
                        **kwargs):
  """Populates an `Experiment` object.
1041

1042
1043
  EXPERIMENT CLASS IS DEPRECATED. Please switch to
  tf.estimator.train_and_evaluate. As an example, see model_main.py.
1044

1045
1046
1047
1048
1049
1050
1051
1052
1053
1054
1055
1056
1057
1058
1059
1060
1061
1062
1063
1064
1065
1066
1067
1068
1069
1070
1071
1072
1073
1074
1075
1076
1077
1078
  Args:
    run_config: A `RunConfig`.
    hparams: A `HParams`.
    pipeline_config_path: A path to a pipeline config file.
    train_steps: Number of training steps. If None, the number of training steps
      is set from the `TrainConfig` proto.
    eval_steps: Number of evaluation steps per evaluation cycle. If None, the
      number of evaluation steps is set from the `EvalConfig` proto.
    model_fn_creator: A function that creates a `model_fn` for `Estimator`.
      Follows the signature:

      * Args:
        * `detection_model_fn`: Function that returns `DetectionModel` instance.
        * `configs`: Dictionary of pipeline config objects.
        * `hparams`: `HParams` object.
      * Returns:
        `model_fn` for `Estimator`.

    **kwargs: Additional keyword arguments for configuration override.

  Returns:
    An `Experiment` that defines all aspects of training, evaluation, and
    export.
  """
  tf.logging.warning('Experiment is being deprecated. Please use '
                     'tf.estimator.train_and_evaluate(). See model_main.py for '
                     'an example.')
  train_and_eval_dict = create_estimator_and_inputs(
      run_config,
      hparams,
      pipeline_config_path,
      train_steps=train_steps,
      eval_steps=eval_steps,
      model_fn_creator=model_fn_creator,
1079
      save_final_config=True,
1080
1081
1082
      **kwargs)
  estimator = train_and_eval_dict['estimator']
  train_input_fn = train_and_eval_dict['train_input_fn']
1083
  eval_input_fns = train_and_eval_dict['eval_input_fns']
1084
1085
1086
1087
  predict_input_fn = train_and_eval_dict['predict_input_fn']
  train_steps = train_and_eval_dict['train_steps']

  export_strategies = [
1088
      contrib_learn.utils.saved_model_export_utils.make_export_strategy(
1089
1090
1091
          serving_input_fn=predict_input_fn)
  ]

1092
  return contrib_learn.Experiment(
1093
1094
      estimator=estimator,
      train_input_fn=train_input_fn,
1095
      eval_input_fn=eval_input_fns[0],
1096
      train_steps=train_steps,
1097
      eval_steps=None,
1098
      export_strategies=export_strategies,
1099
1100
      eval_delay_secs=120,
  )