model_lib.py

# 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.
# ==============================================================================
r"""Constructs model, inputs, and training environment."""

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

import copy
import functools
import os

import tensorflow.compat.v1 as tf
import tensorflow.compat.v2 as tf2
import tf_slim as slim

from object_detection import eval_util
from object_detection import exporter as exporter_lib
from object_detection import inputs
from object_detection.builders import graph_rewriter_builder
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
from object_detection.utils import ops
from object_detection.utils import shape_utils
from object_detection.utils import variables_helper
from object_detection.utils import visualization_utils as vis_utils

# 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

# 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,
    '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,
    'detection_model_fn_base': model_builder.build,
}


def _prepare_groundtruth_for_eval(detection_model, class_agnostic,
                                  max_number_of_boxes):
  """Extracts groundtruth data from detection_model and prepares it for eval.

  Args:
    detection_model: A `DetectionModel` object.
    class_agnostic: Whether the detections are class_agnostic.
    max_number_of_boxes: Max number of groundtruth boxes.

  Returns:
    A tuple of:
    groundtruth: Dictionary with the following fields:
      '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
        groundtruth)
      'groundtruth_is_crowd': [batch_size, num_boxes] bool tensor indicating
        is_crowd annotations (if provided in groundtruth).
      'groundtruth_area': [batch_size, num_boxes] float32 tensor indicating
        the area (in the original absolute coordinates) of annotations (if
        provided in groundtruth).
      'num_groundtruth_boxes': [batch_size] tensor containing the maximum number
        of groundtruth boxes per image..
      'groundtruth_keypoints': [batch_size, num_boxes, num_keypoints, 2] float32
        tensor of keypoints (if provided in groundtruth).
      '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).
      'groundtruth_track_ids_list': [batch_size, num_boxes] int32 tensor
        with track ID for each instance (if provided in groundtruth).
      '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.
      '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.
    class_agnostic: Boolean indicating whether detections are class agnostic.
  """
  input_data_fields = fields.InputDataFields()
  groundtruth_boxes = tf.stack(
      detection_model.groundtruth_lists(fields.BoxListFields.boxes))
  groundtruth_boxes_shape = tf.shape(groundtruth_boxes)
  # For class-agnostic models, groundtruth one-hot encodings collapse to all
  # ones.
  if class_agnostic:
    groundtruth_classes_one_hot = tf.ones(
        [groundtruth_boxes_shape[0], groundtruth_boxes_shape[1], 1])
  else:
    groundtruth_classes_one_hot = tf.stack(
        detection_model.groundtruth_lists(fields.BoxListFields.classes))
  label_id_offset = 1  # Applying label id offset (b/63711816)
  groundtruth_classes = (
      tf.argmax(groundtruth_classes_one_hot, axis=2) + label_id_offset)
  groundtruth = {
      input_data_fields.groundtruth_boxes: groundtruth_boxes,
      input_data_fields.groundtruth_classes: groundtruth_classes
  }

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

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

  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))

  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))

  if detection_model.groundtruth_has_field(
      input_data_fields.groundtruth_verified_neg_classes):
    groundtruth[input_data_fields.groundtruth_verified_neg_classes] = tf.stack(
        detection_model.groundtruth_lists(
            input_data_fields.groundtruth_verified_neg_classes))

  if detection_model.groundtruth_has_field(
      input_data_fields.groundtruth_not_exhaustive_classes):
    groundtruth[
        input_data_fields.groundtruth_not_exhaustive_classes] = tf.stack(
            detection_model.groundtruth_lists(
                input_data_fields.groundtruth_not_exhaustive_classes))

  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))

  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))

  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)

  groundtruth[input_data_fields.num_groundtruth_boxes] = (
      tf.tile([max_number_of_boxes], multiples=[groundtruth_boxes_shape[0]]))
  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
  tensor_dict containing values that are lists of unstacked, unpadded tensors.

  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]

  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
  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.
  """
  unbatched_tensor_dict = {
      key: tf.unstack(tensor) for key, tensor in tensor_dict.items()
  }
  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,
        fields.InputDataFields.groundtruth_keypoint_visibilities,
        fields.InputDataFields.groundtruth_dp_num_points,
        fields.InputDataFields.groundtruth_dp_part_ids,
        fields.InputDataFields.groundtruth_dp_surface_coords,
        fields.InputDataFields.groundtruth_track_ids,
        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

    unbatched_tensor_dict.update(unbatched_unpadded_tensor_dict)

  return unbatched_tensor_dict


def provide_groundtruth(model, labels):
  """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]
  gt_keypoint_visibilities_list = None
  if fields.InputDataFields.groundtruth_keypoint_visibilities in labels:
    gt_keypoint_visibilities_list = labels[
        fields.InputDataFields.groundtruth_keypoint_visibilities]
  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]
  gt_track_ids_list = None
  if fields.InputDataFields.groundtruth_track_ids in labels:
    gt_track_ids_list = labels[
        fields.InputDataFields.groundtruth_track_ids]
  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]
  gt_group_of_list = None
  if fields.InputDataFields.groundtruth_group_of in labels:
    gt_group_of_list = labels[fields.InputDataFields.groundtruth_group_of]
  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]
  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]
  model.provide_groundtruth(
      groundtruth_boxes_list=gt_boxes_list,
      groundtruth_classes_list=gt_classes_list,
      groundtruth_confidences_list=gt_confidences_list,
      groundtruth_labeled_classes=gt_labeled_classes,
      groundtruth_masks_list=gt_masks_list,
      groundtruth_keypoints_list=gt_keypoints_list,
      groundtruth_keypoint_visibilities_list=gt_keypoint_visibilities_list,
      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,
      groundtruth_weights_list=gt_weights_list,
      groundtruth_is_crowd_list=gt_is_crowd_list,
      groundtruth_group_of_list=gt_group_of_list,
      groundtruth_area_list=gt_area_list,
      groundtruth_track_ids_list=gt_track_ids_list,
      groundtruth_verified_neg_classes=gt_verified_neg_classes,
      groundtruth_not_exhaustive_classes=gt_not_exhaustive_classes)


def create_model_fn(detection_model_fn, configs, hparams=None, use_tpu=False,
                    postprocess_on_cpu=False):
  """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.
    postprocess_on_cpu: When use_tpu and postprocess_on_cpu is true, postprocess
        is scheduled on the host cpu.

  Returns:
    `model_fn` for `Estimator`.
  """
  train_config = configs['train_config']
  eval_input_config = configs['eval_input_config']
  eval_config = configs['eval_config']

  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

    # Make sure to set the Keras learning phase. True during training,
    # False for inference.
    tf.keras.backend.set_learning_phase(is_training)
    # 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()
      tf2.keras.mixed_precision.experimental.set_policy(
          'mixed_bfloat16')
    detection_model = detection_model_fn(
        is_training=is_training, add_summaries=(not use_tpu))
    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:
      # 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())
      unpad_groundtruth_tensors = boxes_shape[1] is not None and not use_tpu
      labels = unstack_batch(
          labels, unpad_groundtruth_tensors=unpad_groundtruth_tensors)

    if mode in (tf.estimator.ModeKeys.TRAIN, tf.estimator.ModeKeys.EVAL):
      provide_groundtruth(detection_model, labels)

    preprocessed_images = features[fields.InputDataFields.image]

    side_inputs = detection_model.get_side_inputs(features)

    if use_tpu and train_config.use_bfloat16:
      with tf.tpu.bfloat16_scope():
        prediction_dict = detection_model.predict(
            preprocessed_images,
            features[fields.InputDataFields.true_image_shape], **side_inputs)
        prediction_dict = ops.bfloat16_to_float32_nested(prediction_dict)
    else:
      prediction_dict = detection_model.predict(
          preprocessed_images,
          features[fields.InputDataFields.true_image_shape], **side_inputs)

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

    if mode in (tf.estimator.ModeKeys.EVAL, tf.estimator.ModeKeys.PREDICT):
      if use_tpu and postprocess_on_cpu:
        detections = tf.tpu.outside_compilation(
            postprocess_wrapper,
            (prediction_dict,
             features[fields.InputDataFields.true_image_shape]))
      else:
        detections = postprocess_wrapper((
            prediction_dict,
            features[fields.InputDataFields.true_image_shape]))

    if mode == tf.estimator.ModeKeys.TRAIN:
      load_pretrained = hparams.load_pretrained if hparams else False
      if train_config.fine_tune_checkpoint and load_pretrained:
        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'
        asg_map = detection_model.restore_map(
            fine_tune_checkpoint_type=train_config.fine_tune_checkpoint_type,
            load_all_detection_checkpoint_vars=(
                train_config.load_all_detection_checkpoint_vars))
        available_var_map = (
            variables_helper.get_variables_available_in_checkpoint(
                asg_map,
                train_config.fine_tune_checkpoint,
                include_global_step=False))
        if use_tpu:

          def tpu_scaffold():
            tf.train.init_from_checkpoint(train_config.fine_tune_checkpoint,
                                          available_var_map)
            return tf.train.Scaffold()

          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):
      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

      if 'graph_rewriter_config' in configs:
        graph_rewriter_fn = graph_rewriter_builder.build(
            configs['graph_rewriter_config'], is_training=is_training)
        graph_rewriter_fn()

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

    if mode == tf.estimator.ModeKeys.TRAIN:
      if use_tpu:
        training_optimizer = tf.tpu.CrossShardOptimizer(training_optimizer)

      # Optionally freeze some layers by setting their gradients to be zero.
      trainable_variables = None
      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)
      trainable_variables = slim.filter_variables(
          tf.trainable_variables(),
          include_patterns=include_variables,
          exclude_patterns=exclude_variables)

      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
      if train_config.summarize_gradients:
        summaries = ['gradients', 'gradient_norm', 'global_gradient_norm']
      train_op = slim.optimizers.optimize_loss(
          loss=total_loss,
          global_step=global_step,
          learning_rate=None,
          clip_gradients=clip_gradients_value,
          optimizer=training_optimizer,
          update_ops=detection_model.updates(),
          variables=trainable_variables,
          summaries=summaries,
          name='')  # Preventing scope prefix on all variables.

    if mode == tf.estimator.ModeKeys.PREDICT:
      exported_output = exporter_lib.add_output_tensor_nodes(detections)
      export_outputs = {
          tf.saved_model.signature_constants.PREDICT_METHOD_NAME:
              tf.estimator.export.PredictOutput(exported_output)
      }

    eval_metric_ops = None
    scaffold = None
    if mode == tf.estimator.ModeKeys.EVAL:
      class_agnostic = (
          fields.DetectionResultFields.detection_classes not in detections)
      groundtruth = _prepare_groundtruth_for_eval(
          detection_model, class_agnostic,
          eval_input_config.max_number_of_boxes)
      use_original_images = fields.InputDataFields.original_image in features
      if use_original_images:
        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]
      else:
        eval_images = features[fields.InputDataFields.image]
        true_image_shapes = None
        original_image_spatial_shapes = None

      eval_dict = eval_util.result_dict_for_batched_example(
          eval_images,
          features[inputs.HASH_KEY],
          detections,
          groundtruth,
          class_agnostic=class_agnostic,
          scale_to_absolute=True,
          original_image_spatial_shapes=original_image_spatial_shapes,
          true_image_shapes=true_image_shapes)

      if fields.InputDataFields.image_additional_channels in features:
        eval_dict[fields.InputDataFields.image_additional_channels] = features[
            fields.InputDataFields.image_additional_channels]

      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)
      vis_metric_ops = None
      if not use_tpu and use_original_images:
        keypoint_edges = [
            (kp.start, kp.end) for kp in eval_config.keypoint_edge]

        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,
            use_normalized_coordinates=False,
            keypoint_edges=keypoint_edges or None)
        vis_metric_ops = eval_metric_op_vis.get_estimator_eval_metric_ops(
            eval_dict)

      # Eval metrics on a single example.
      eval_metric_ops = eval_util.get_eval_metric_ops_for_evaluators(
          eval_config, list(category_index.values()), eval_dict)
      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())
      if vis_metric_ops is not None:
        eval_metric_ops.update(vis_metric_ops)
      eval_metric_ops = {str(k): v for k, v in eval_metric_ops.items()}

      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)

    # EVAL executes on CPU, so use regular non-TPU EstimatorSpec.
    if use_tpu and mode != tf.estimator.ModeKeys.EVAL:
      return tf.estimator.tpu.TPUEstimatorSpec(
          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:
      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)
      return tf.estimator.EstimatorSpec(
          mode=mode,
          predictions=detections,
          loss=total_loss,
          train_op=train_op,
          eval_metric_ops=eval_metric_ops,
          export_outputs=export_outputs,
          scaffold=scaffold)

  return model_fn


def create_estimator_and_inputs(run_config,
                                hparams=None,
                                pipeline_config_path=None,
                                config_override=None,
                                train_steps=None,
                                sample_1_of_n_eval_examples=1,
                                sample_1_of_n_eval_on_train_examples=1,
                                model_fn_creator=create_model_fn,
                                use_tpu_estimator=False,
                                use_tpu=False,
                                num_shards=1,
                                params=None,
                                override_eval_num_epochs=True,
                                save_final_config=False,
                                postprocess_on_cpu=False,
                                export_to_tpu=None,
                                **kwargs):
  """Creates `Estimator`, input functions, and steps.

  Args:
    run_config: A `RunConfig`.
    hparams: (optional) A `HParams`.
    pipeline_config_path: A path to a pipeline config file.
    config_override: A pipeline_pb2.TrainEvalPipelineConfig text proto to
      override the config from `pipeline_config_path`.
    train_steps: Number of training steps. If None, the number of training steps
      is set from the `TrainConfig` proto.
    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.
    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`.

    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.
    override_eval_num_epochs: Whether to overwrite the number of epochs to 1 for
      eval_input.
    save_final_config: Whether to save final config (obtained after applying
      overrides) to `estimator.model_dir`.
    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.
    **kwargs: Additional keyword arguments for configuration override.

  Returns:
    A dictionary with the following fields:
    'estimator': An `Estimator` or `TPUEstimator`.
    'train_input_fn': A training input function.
    'eval_input_fns': A list of all evaluation input functions.
    'eval_input_names': A list of names for each evaluation input.
    'eval_on_train_input_fn': An evaluation-on-train input function.
    'predict_input_fn': A prediction input function.
    'train_steps': Number of training steps. Either directly from input or from
      configuration.
  """
  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']
  create_pipeline_proto_from_configs = MODEL_BUILD_UTIL_MAP[
      'create_pipeline_proto_from_configs']
  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']
  detection_model_fn_base = MODEL_BUILD_UTIL_MAP['detection_model_fn_base']

  configs = get_configs_from_pipeline_file(
      pipeline_config_path, config_override=config_override)
  kwargs.update({
      'train_steps': train_steps,
      'use_bfloat16': configs['train_config'].use_bfloat16 and use_tpu
  })
  if sample_1_of_n_eval_examples >= 1:
    kwargs.update({
        'sample_1_of_n_eval_examples': sample_1_of_n_eval_examples
    })
  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.')
  configs = merge_external_params_with_configs(
      configs, hparams, kwargs_dict=kwargs)
  model_config = configs['model']
  train_config = configs['train_config']
  train_input_config = configs['train_input_config']
  eval_config = configs['eval_config']
  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

  # 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

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

  # Create the input functions for TRAIN/EVAL/PREDICT.
  train_input_fn = create_train_input_fn(
      train_config=train_config,
      train_input_config=train_input_config,
      model_config=model_config)
  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
  ]
  eval_on_train_input_fn = create_eval_input_fn(
      eval_config=eval_config,
      eval_input_config=eval_on_train_input_config,
      model_config=model_config)
  predict_input_fn = create_predict_input_fn(
      model_config=model_config, predict_input_config=eval_input_configs[0])

  # Read export_to_tpu from hparams if not passed.
  if export_to_tpu is None and hparams is not None:
    export_to_tpu = hparams.get('export_to_tpu', False)
  tf.logging.info('create_estimator_and_inputs: use_tpu %s, export_to_tpu %s',
                  use_tpu, export_to_tpu)
  model_fn = model_fn_creator(detection_model_fn, configs, hparams, use_tpu,
                              postprocess_on_cpu)
  if use_tpu_estimator:
    estimator = tf.estimator.tpu.TPUEstimator(
        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,
        export_to_tpu=export_to_tpu,
        eval_on_tpu=False,  # Eval runs on CPU, so disable eval on TPU
        params=params if params else {})
  else:
    estimator = tf.estimator.Estimator(model_fn=model_fn, config=run_config)

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

  return dict(
      estimator=estimator,
      train_input_fn=train_input_fn,
      eval_input_fns=eval_input_fns,
      eval_input_names=eval_input_names,
      eval_on_train_input_fn=eval_on_train_input_fn,
      predict_input_fn=predict_input_fn,
      train_steps=train_steps)


def create_train_and_eval_specs(train_input_fn,
                                eval_input_fns,
                                eval_on_train_input_fn,
                                predict_input_fn,
                                train_steps,
                                eval_on_train_data=False,
                                final_exporter_name='Servo',
                                eval_spec_names=None):
  """Creates a `TrainSpec` and `EvalSpec`s.

  Args:
    train_input_fn: Function that produces features and labels on train data.
    eval_input_fns: A list of functions that produce features and labels on eval
      data.
    eval_on_train_input_fn: Function that produces features and labels for
      evaluation on train data.
    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`.
    eval_spec_names: A list of string names for each `EvalSpec`.

  Returns:
    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.
  """
  train_spec = tf.estimator.TrainSpec(
      input_fn=train_input_fn, max_steps=train_steps)

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

  eval_specs = []
  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)
    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))

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

  return train_spec, eval_specs


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):
  """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.
    max_retries: Maximum number of times to retry the evaluation on encountering
      a tf.errors.InvalidArgumentError. If negative, will always retry the
      evaluation.
  """

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

  for ckpt in tf.train.checkpoints_iterator(
      model_dir, min_interval_secs=180, timeout=None,
      timeout_fn=terminate_eval):

    tf.logging.info('Starting Evaluation.')
    try:
      eval_results = _evaluate_checkpoint(
          estimator=estimator,
          input_fn=input_fn,
          checkpoint_path=ckpt,
          name=name,
          max_retries=max_retries)
      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)


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.

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

  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,
      save_final_config=True,
      **kwargs)
  estimator = train_and_eval_dict['estimator']
  train_input_fn = train_and_eval_dict['train_input_fn']
  eval_input_fns = train_and_eval_dict['eval_input_fns']
  predict_input_fn = train_and_eval_dict['predict_input_fn']
  train_steps = train_and_eval_dict['train_steps']

  export_strategies = [
      contrib_learn.utils.saved_model_export_utils.make_export_strategy(
          serving_input_fn=predict_input_fn)
  ]

  return contrib_learn.Experiment(
      estimator=estimator,
      train_input_fn=train_input_fn,
      eval_input_fn=eval_input_fns[0],
      train_steps=train_steps,
      eval_steps=None,
      export_strategies=export_strategies,
      eval_delay_secs=120,
  )