Python object_detection.meta_architectures.faster_rcnn_meta_arch.FasterRCNNFeatureExtractor() Examples

The following are 30 code examples of object_detection.meta_architectures.faster_rcnn_meta_arch.FasterRCNNFeatureExtractor(). You can vote up the ones you like or vote down the ones you don't like, and go to the original project or source file by following the links above each example. You may also want to check out all available functions/classes of the module object_detection.meta_architectures.faster_rcnn_meta_arch , or try the search function .
Example #1
Source File: faster_rcnn_nas_feature_extractor.py    From Elphas with Apache License 2.0 5 votes vote down vote up 
def restore_from_classification_checkpoint_fn(
      self,
      first_stage_feature_extractor_scope,
      second_stage_feature_extractor_scope):
    """Returns a map of variables to load from a foreign checkpoint.

    Note that this overrides the default implementation in
    faster_rcnn_meta_arch.FasterRCNNFeatureExtractor which does not work for
    NASNet-A checkpoints.

    Args:
      first_stage_feature_extractor_scope: A scope name for the first stage
        feature extractor.
      second_stage_feature_extractor_scope: A scope name for the second stage
        feature extractor.

    Returns:
      A dict mapping variable names (to load from a checkpoint) to variables in
      the model graph.
    """
    # Note that the NAS checkpoint only contains the moving average version of
    # the Variables so we need to generate an appropriate dictionary mapping.
    variables_to_restore = {}
    for variable in tf.global_variables():
      if variable.op.name.startswith(
          first_stage_feature_extractor_scope):
        var_name = variable.op.name.replace(
            first_stage_feature_extractor_scope + '/', '')
        var_name += '/ExponentialMovingAverage'
        variables_to_restore[var_name] = variable
      if variable.op.name.startswith(
          second_stage_feature_extractor_scope):
        var_name = variable.op.name.replace(
            second_stage_feature_extractor_scope + '/', '')
        var_name += '/ExponentialMovingAverage'
        variables_to_restore[var_name] = variable
    return variables_to_restore
Example #2
Source File: model_builder.py    From MAX-Object-Detector with Apache License 2.0 5 votes vote down vote up 
def _build_faster_rcnn_feature_extractor(
    feature_extractor_config, is_training, reuse_weights=None,
    inplace_batchnorm_update=False):
  """Builds a faster_rcnn_meta_arch.FasterRCNNFeatureExtractor based on config.

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

  Returns:
    faster_rcnn_meta_arch.FasterRCNNFeatureExtractor based on config.

  Raises:
    ValueError: On invalid feature extractor type.
  """
  if inplace_batchnorm_update:
    raise ValueError('inplace batchnorm updates not supported.')
  feature_type = feature_extractor_config.type
  first_stage_features_stride = (
      feature_extractor_config.first_stage_features_stride)
  batch_norm_trainable = feature_extractor_config.batch_norm_trainable

  if feature_type not in FASTER_RCNN_FEATURE_EXTRACTOR_CLASS_MAP:
    raise ValueError('Unknown Faster R-CNN feature_extractor: {}'.format(
        feature_type))
  feature_extractor_class = FASTER_RCNN_FEATURE_EXTRACTOR_CLASS_MAP[
      feature_type]
  return feature_extractor_class(
      is_training, first_stage_features_stride,
      batch_norm_trainable, reuse_weights)
Example #3
Source File: model_builder.py    From AniSeg with Apache License 2.0 5 votes vote down vote up 
def _build_faster_rcnn_feature_extractor(
    feature_extractor_config, is_training, reuse_weights=None):
  """Builds a faster_rcnn_meta_arch.FasterRCNNFeatureExtractor based on config.

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

  Returns:
    faster_rcnn_meta_arch.FasterRCNNFeatureExtractor based on config.

  Raises:
    ValueError: On invalid feature extractor type.
  """
  feature_type = feature_extractor_config.type
  first_stage_features_stride = (
      feature_extractor_config.first_stage_features_stride)
  batch_norm_trainable = feature_extractor_config.batch_norm_trainable

  if feature_type not in FASTER_RCNN_FEATURE_EXTRACTOR_CLASS_MAP:
    raise ValueError('Unknown Faster R-CNN feature_extractor: {}'.format(
        feature_type))
  feature_extractor_class = FASTER_RCNN_FEATURE_EXTRACTOR_CLASS_MAP[
      feature_type]
  return feature_extractor_class(
      is_training, first_stage_features_stride,
      batch_norm_trainable, reuse_weights)
Example #4
Source File: faster_rcnn_nas_feature_extractor.py    From object_detection_with_tensorflow with MIT License 5 votes vote down vote up 
def restore_from_classification_checkpoint_fn(
      self,
      first_stage_feature_extractor_scope,
      second_stage_feature_extractor_scope):
    """Returns a map of variables to load from a foreign checkpoint.

    Note that this overrides the default implementation in
    faster_rcnn_meta_arch.FasterRCNNFeatureExtractor which does not work for
    NASNet-A checkpoints.

    Args:
      first_stage_feature_extractor_scope: A scope name for the first stage
        feature extractor.
      second_stage_feature_extractor_scope: A scope name for the second stage
        feature extractor.

    Returns:
      A dict mapping variable names (to load from a checkpoint) to variables in
      the model graph.
    """
    # Note that the NAS checkpoint only contains the moving average version of
    # the Variables so we need to generate an appropriate dictionary mapping.
    variables_to_restore = {}
    for variable in tf.global_variables():
      if variable.op.name.startswith(
          first_stage_feature_extractor_scope):
        var_name = variable.op.name.replace(
            first_stage_feature_extractor_scope + '/', '')
        var_name += '/ExponentialMovingAverage'
        variables_to_restore[var_name] = variable
      if variable.op.name.startswith(
          second_stage_feature_extractor_scope):
        var_name = variable.op.name.replace(
            second_stage_feature_extractor_scope + '/', '')
        var_name += '/ExponentialMovingAverage'
        variables_to_restore[var_name] = variable
    return variables_to_restore
Example #5
Source File: model_builder.py    From object_detection_with_tensorflow with MIT License 5 votes vote down vote up 
def _build_faster_rcnn_feature_extractor(
    feature_extractor_config, is_training, reuse_weights=None):
  """Builds a faster_rcnn_meta_arch.FasterRCNNFeatureExtractor based on config.

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

  Returns:
    faster_rcnn_meta_arch.FasterRCNNFeatureExtractor based on config.

  Raises:
    ValueError: On invalid feature extractor type.
  """
  feature_type = feature_extractor_config.type
  first_stage_features_stride = (
      feature_extractor_config.first_stage_features_stride)
  batch_norm_trainable = feature_extractor_config.batch_norm_trainable

  if feature_type not in FASTER_RCNN_FEATURE_EXTRACTOR_CLASS_MAP:
    raise ValueError('Unknown Faster R-CNN feature_extractor: {}'.format(
        feature_type))
  feature_extractor_class = FASTER_RCNN_FEATURE_EXTRACTOR_CLASS_MAP[
      feature_type]
  return feature_extractor_class(
      is_training, first_stage_features_stride,
      batch_norm_trainable, reuse_weights)
Example #6
Source File: faster_rcnn_nas_feature_extractor.py    From MAX-Object-Detector with Apache License 2.0 5 votes vote down vote up 
def restore_from_classification_checkpoint_fn(
      self,
      first_stage_feature_extractor_scope,
      second_stage_feature_extractor_scope):
    """Returns a map of variables to load from a foreign checkpoint.

    Note that this overrides the default implementation in
    faster_rcnn_meta_arch.FasterRCNNFeatureExtractor which does not work for
    NASNet-A checkpoints.

    Args:
      first_stage_feature_extractor_scope: A scope name for the first stage
        feature extractor.
      second_stage_feature_extractor_scope: A scope name for the second stage
        feature extractor.

    Returns:
      A dict mapping variable names (to load from a checkpoint) to variables in
      the model graph.
    """
    # Note that the NAS checkpoint only contains the moving average version of
    # the Variables so we need to generate an appropriate dictionary mapping.
    variables_to_restore = {}
    for variable in tf.global_variables():
      if variable.op.name.startswith(
          first_stage_feature_extractor_scope):
        var_name = variable.op.name.replace(
            first_stage_feature_extractor_scope + '/', '')
        var_name += '/ExponentialMovingAverage'
        variables_to_restore[var_name] = variable
      if variable.op.name.startswith(
          second_stage_feature_extractor_scope):
        var_name = variable.op.name.replace(
            second_stage_feature_extractor_scope + '/', '')
        var_name += '/ExponentialMovingAverage'
        variables_to_restore[var_name] = variable
    return variables_to_restore
Example #7
Source File: model_builder.py    From g-tensorflow-models with Apache License 2.0 5 votes vote down vote up 
def _build_faster_rcnn_feature_extractor(
    feature_extractor_config, is_training, reuse_weights=None,
    inplace_batchnorm_update=False):
  """Builds a faster_rcnn_meta_arch.FasterRCNNFeatureExtractor based on config.

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

  Returns:
    faster_rcnn_meta_arch.FasterRCNNFeatureExtractor based on config.

  Raises:
    ValueError: On invalid feature extractor type.
  """
  if inplace_batchnorm_update:
    raise ValueError('inplace batchnorm updates not supported.')
  feature_type = feature_extractor_config.type
  first_stage_features_stride = (
      feature_extractor_config.first_stage_features_stride)
  batch_norm_trainable = feature_extractor_config.batch_norm_trainable

  if feature_type not in FASTER_RCNN_FEATURE_EXTRACTOR_CLASS_MAP:
    raise ValueError('Unknown Faster R-CNN feature_extractor: {}'.format(
        feature_type))
  feature_extractor_class = FASTER_RCNN_FEATURE_EXTRACTOR_CLASS_MAP[
      feature_type]
  return feature_extractor_class(
      is_training, first_stage_features_stride,
      batch_norm_trainable, reuse_weights)
Example #8
Source File: faster_rcnn_nas_feature_extractor.py    From models with Apache License 2.0 5 votes vote down vote up 
def restore_from_classification_checkpoint_fn(
      self,
      first_stage_feature_extractor_scope,
      second_stage_feature_extractor_scope):
    """Returns a map of variables to load from a foreign checkpoint.

    Note that this overrides the default implementation in
    faster_rcnn_meta_arch.FasterRCNNFeatureExtractor which does not work for
    NASNet-A checkpoints.

    Args:
      first_stage_feature_extractor_scope: A scope name for the first stage
        feature extractor.
      second_stage_feature_extractor_scope: A scope name for the second stage
        feature extractor.

    Returns:
      A dict mapping variable names (to load from a checkpoint) to variables in
      the model graph.
    """
    # Note that the NAS checkpoint only contains the moving average version of
    # the Variables so we need to generate an appropriate dictionary mapping.
    variables_to_restore = {}
    for variable in variables_helper.get_global_variables_safely():
      if variable.op.name.startswith(
          first_stage_feature_extractor_scope):
        var_name = variable.op.name.replace(
            first_stage_feature_extractor_scope + '/', '')
        var_name += '/ExponentialMovingAverage'
        variables_to_restore[var_name] = variable
      if variable.op.name.startswith(
          second_stage_feature_extractor_scope):
        var_name = variable.op.name.replace(
            second_stage_feature_extractor_scope + '/', '')
        var_name += '/ExponentialMovingAverage'
        variables_to_restore[var_name] = variable
    return variables_to_restore
Example #9
Source File: faster_rcnn_pnas_feature_extractor.py    From models with Apache License 2.0 5 votes vote down vote up 
def restore_from_classification_checkpoint_fn(
      self,
      first_stage_feature_extractor_scope,
      second_stage_feature_extractor_scope):
    """Returns a map of variables to load from a foreign checkpoint.

    Note that this overrides the default implementation in
    faster_rcnn_meta_arch.FasterRCNNFeatureExtractor which does not work for
    PNASNet checkpoints.

    Args:
      first_stage_feature_extractor_scope: A scope name for the first stage
        feature extractor.
      second_stage_feature_extractor_scope: A scope name for the second stage
        feature extractor.

    Returns:
      A dict mapping variable names (to load from a checkpoint) to variables in
      the model graph.
    """
    variables_to_restore = {}
    for variable in variables_helper.get_global_variables_safely():
      if variable.op.name.startswith(
          first_stage_feature_extractor_scope):
        var_name = variable.op.name.replace(
            first_stage_feature_extractor_scope + '/', '')
        var_name += '/ExponentialMovingAverage'
        variables_to_restore[var_name] = variable
      if variable.op.name.startswith(
          second_stage_feature_extractor_scope):
        var_name = variable.op.name.replace(
            second_stage_feature_extractor_scope + '/', '')
        var_name += '/ExponentialMovingAverage'
        variables_to_restore[var_name] = variable
    return variables_to_restore
Example #10
Source File: model_builder.py    From models with Apache License 2.0 5 votes vote down vote up 
def _build_faster_rcnn_feature_extractor(
    feature_extractor_config, is_training, reuse_weights=True,
    inplace_batchnorm_update=False):
  """Builds a faster_rcnn_meta_arch.FasterRCNNFeatureExtractor based on config.

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

  Returns:
    faster_rcnn_meta_arch.FasterRCNNFeatureExtractor based on config.

  Raises:
    ValueError: On invalid feature extractor type.
  """
  if inplace_batchnorm_update:
    raise ValueError('inplace batchnorm updates not supported.')
  feature_type = feature_extractor_config.type
  first_stage_features_stride = (
      feature_extractor_config.first_stage_features_stride)
  batch_norm_trainable = feature_extractor_config.batch_norm_trainable

  if feature_type not in FASTER_RCNN_FEATURE_EXTRACTOR_CLASS_MAP:
    raise ValueError('Unknown Faster R-CNN feature_extractor: {}'.format(
        feature_type))
  feature_extractor_class = FASTER_RCNN_FEATURE_EXTRACTOR_CLASS_MAP[
      feature_type]
  return feature_extractor_class(
      is_training, first_stage_features_stride,
      batch_norm_trainable, reuse_weights=reuse_weights)
Example #11
Source File: model_builder.py    From motion-rcnn with MIT License 5 votes vote down vote up 
def _build_faster_rcnn_feature_extractor(
    feature_extractor_config, is_training, reuse_weights=None):
  """Builds a faster_rcnn_meta_arch.FasterRCNNFeatureExtractor based on config.

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

  Returns:
    faster_rcnn_meta_arch.FasterRCNNFeatureExtractor based on config.

  Raises:
    ValueError: On invalid feature extractor type.
  """
  feature_type = feature_extractor_config.type
  first_stage_features_stride = (
      feature_extractor_config.first_stage_features_stride)

  if feature_type not in FASTER_RCNN_FEATURE_EXTRACTOR_CLASS_MAP:
    raise ValueError('Unknown Faster R-CNN feature_extractor: {}'.format(
        feature_type))
  feature_extractor_class = FASTER_RCNN_FEATURE_EXTRACTOR_CLASS_MAP[
      feature_type]
  return feature_extractor_class(
      is_training, first_stage_features_stride, reuse_weights)
Example #12
Source File: faster_rcnn_inception_resnet_v2_feature_extractor.py    From mtl-ssl with Apache License 2.0 5 votes vote down vote up 
def mtl_restore_from_classification_checkpoint_fn(self, scope_name):
    """Returns a map of variables to load from a foreign checkpoint.

    Note that this overrides the default implementation in
    faster_rcnn_meta_arch.FasterRCNNFeatureExtractor which does not work for
    InceptionResnetV2 checkpoints.

    TODO: revisit whether it's possible to force the
    `Repeat` namescope as created in `_extract_box_classifier_features` to
    start counting at 2 (e.g. `Repeat_2`) so that the default restore_fn can
    be used.

    Args:
      first_stage_feature_extractor_scope: A scope name for the first stage
        feature extractor.
      second_stage_feature_extractor_scope: A scope name for the second stage
        feature extractor.

    Returns:
      A dict mapping variable names (to load from a checkpoint) to variables in
      the model graph.
    """

    variables_to_restore = {}
    for variable in tf.global_variables():
      if variable.op.name.startswith(
          scope_name):
        var_name = variable.op.name.replace(
          scope_name
            + '/InceptionResnetV2/Repeat', 'InceptionResnetV2/Repeat_2')
        var_name = var_name.replace(
          scope_name + '/', '')
        variables_to_restore[var_name] = variable
    return variables_to_restore
Example #13
Source File: model_builder.py    From mtl-ssl with Apache License 2.0 5 votes vote down vote up 
def _build_faster_rcnn_feature_extractor(
    feature_extractor_config, is_training, reuse_weights, **kwargs):
  """Builds a faster_rcnn_meta_arch.FasterRCNNFeatureExtractor based on config.

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

  Returns:
    faster_rcnn_meta_arch.FasterRCNNFeatureExtractor based on config.

  Raises:
    ValueError: On invalid feature extractor type.
  """
  feature_type = feature_extractor_config.type
  first_stage_features_stride = (
      feature_extractor_config.first_stage_features_stride)

  if feature_type not in FASTER_RCNN_FEATURE_EXTRACTOR_CLASS_MAP:
    raise ValueError('Unknown Faster R-CNN feature_extractor: {}'.format(
        feature_type))
  feature_extractor_class = FASTER_RCNN_FEATURE_EXTRACTOR_CLASS_MAP[
      feature_type]
  return feature_extractor_class(
      is_training, first_stage_features_stride, reuse_weights, **kwargs)
Example #14
Source File: faster_rcnn_pnas_feature_extractor.py    From g-tensorflow-models with Apache License 2.0 5 votes vote down vote up 
def restore_from_classification_checkpoint_fn(
      self,
      first_stage_feature_extractor_scope,
      second_stage_feature_extractor_scope):
    """Returns a map of variables to load from a foreign checkpoint.

    Note that this overrides the default implementation in
    faster_rcnn_meta_arch.FasterRCNNFeatureExtractor which does not work for
    PNASNet checkpoints.

    Args:
      first_stage_feature_extractor_scope: A scope name for the first stage
        feature extractor.
      second_stage_feature_extractor_scope: A scope name for the second stage
        feature extractor.

    Returns:
      A dict mapping variable names (to load from a checkpoint) to variables in
      the model graph.
    """
    variables_to_restore = {}
    for variable in tf.global_variables():
      if variable.op.name.startswith(
          first_stage_feature_extractor_scope):
        var_name = variable.op.name.replace(
            first_stage_feature_extractor_scope + '/', '')
        var_name += '/ExponentialMovingAverage'
        variables_to_restore[var_name] = variable
      if variable.op.name.startswith(
          second_stage_feature_extractor_scope):
        var_name = variable.op.name.replace(
            second_stage_feature_extractor_scope + '/', '')
        var_name += '/ExponentialMovingAverage'
        variables_to_restore[var_name] = variable
    return variables_to_restore
Example #15
Source File: model_builder.py    From Elphas with Apache License 2.0 5 votes vote down vote up 
def _build_faster_rcnn_feature_extractor(
    feature_extractor_config, is_training, reuse_weights=None):
  """Builds a faster_rcnn_meta_arch.FasterRCNNFeatureExtractor based on config.

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

  Returns:
    faster_rcnn_meta_arch.FasterRCNNFeatureExtractor based on config.

  Raises:
    ValueError: On invalid feature extractor type.
  """
  feature_type = feature_extractor_config.type
  first_stage_features_stride = (
      feature_extractor_config.first_stage_features_stride)
  batch_norm_trainable = feature_extractor_config.batch_norm_trainable

  if feature_type not in FASTER_RCNN_FEATURE_EXTRACTOR_CLASS_MAP:
    raise ValueError('Unknown Faster R-CNN feature_extractor: {}'.format(
        feature_type))
  feature_extractor_class = FASTER_RCNN_FEATURE_EXTRACTOR_CLASS_MAP[
      feature_type]
  return feature_extractor_class(
      is_training, first_stage_features_stride,
      batch_norm_trainable, reuse_weights)
Example #16
Source File: faster_rcnn_nas_feature_extractor.py    From object_detection_with_tensorflow with MIT License 5 votes vote down vote up 
def restore_from_classification_checkpoint_fn(
      self,
      first_stage_feature_extractor_scope,
      second_stage_feature_extractor_scope):
    """Returns a map of variables to load from a foreign checkpoint.

    Note that this overrides the default implementation in
    faster_rcnn_meta_arch.FasterRCNNFeatureExtractor which does not work for
    NASNet-A checkpoints.

    Args:
      first_stage_feature_extractor_scope: A scope name for the first stage
        feature extractor.
      second_stage_feature_extractor_scope: A scope name for the second stage
        feature extractor.

    Returns:
      A dict mapping variable names (to load from a checkpoint) to variables in
      the model graph.
    """
    # Note that the NAS checkpoint only contains the moving average version of
    # the Variables so we need to generate an appropriate dictionary mapping.
    variables_to_restore = {}
    for variable in tf.global_variables():
      if variable.op.name.startswith(
          first_stage_feature_extractor_scope):
        var_name = variable.op.name.replace(
            first_stage_feature_extractor_scope + '/', '')
        var_name += '/ExponentialMovingAverage'
        variables_to_restore[var_name] = variable
      if variable.op.name.startswith(
          second_stage_feature_extractor_scope):
        var_name = variable.op.name.replace(
            second_stage_feature_extractor_scope + '/', '')
        var_name += '/ExponentialMovingAverage'
        variables_to_restore[var_name] = variable
    return variables_to_restore
Example #17
Source File: model_builder.py    From MBMD with MIT License 5 votes vote down vote up 
def _build_faster_rcnn_feature_extractor(
    feature_extractor_config, is_training, reuse_weights=None):
  """Builds a faster_rcnn_meta_arch.FasterRCNNFeatureExtractor based on config.

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

  Returns:
    faster_rcnn_meta_arch.FasterRCNNFeatureExtractor based on config.

  Raises:
    ValueError: On invalid feature extractor type.
  """
  feature_type = feature_extractor_config.type
  first_stage_features_stride = (
      feature_extractor_config.first_stage_features_stride)

  if feature_type not in FASTER_RCNN_FEATURE_EXTRACTOR_CLASS_MAP:
    raise ValueError('Unknown Faster R-CNN feature_extractor: {}'.format(
        feature_type))
  feature_extractor_class = FASTER_RCNN_FEATURE_EXTRACTOR_CLASS_MAP[
      feature_type]
  return feature_extractor_class(
      is_training, first_stage_features_stride, reuse_weights)
Example #18
Source File: model_builder.py    From object_detection_kitti with Apache License 2.0 5 votes vote down vote up 
def _build_faster_rcnn_feature_extractor(
    feature_extractor_config, is_training, reuse_weights=None):
  """Builds a faster_rcnn_meta_arch.FasterRCNNFeatureExtractor based on config.

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

  Returns:
    faster_rcnn_meta_arch.FasterRCNNFeatureExtractor based on config.

  Raises:
    ValueError: On invalid feature extractor type.
  """
  feature_type = feature_extractor_config.type
  first_stage_features_stride = (
      feature_extractor_config.first_stage_features_stride)

  if feature_type not in FASTER_RCNN_FEATURE_EXTRACTOR_CLASS_MAP:
    raise ValueError('Unknown Faster R-CNN feature_extractor: {}'.format(
        feature_type))
  feature_extractor_class = FASTER_RCNN_FEATURE_EXTRACTOR_CLASS_MAP[
      feature_type]
  return feature_extractor_class(
      is_training, first_stage_features_stride, reuse_weights)
Example #19
Source File: model_builder.py    From Live-feed-object-device-identification-using-Tensorflow-and-OpenCV with Apache License 2.0 5 votes vote down vote up 
def _build_faster_rcnn_feature_extractor(
    feature_extractor_config, is_training, reuse_weights=None,
    inplace_batchnorm_update=False):
  """Builds a faster_rcnn_meta_arch.FasterRCNNFeatureExtractor based on config.

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

  Returns:
    faster_rcnn_meta_arch.FasterRCNNFeatureExtractor based on config.

  Raises:
    ValueError: On invalid feature extractor type.
  """
  if inplace_batchnorm_update:
    raise ValueError('inplace batchnorm updates not supported.')
  feature_type = feature_extractor_config.type
  first_stage_features_stride = (
      feature_extractor_config.first_stage_features_stride)
  batch_norm_trainable = feature_extractor_config.batch_norm_trainable

  if feature_type not in FASTER_RCNN_FEATURE_EXTRACTOR_CLASS_MAP:
    raise ValueError('Unknown Faster R-CNN feature_extractor: {}'.format(
        feature_type))
  feature_extractor_class = FASTER_RCNN_FEATURE_EXTRACTOR_CLASS_MAP[
      feature_type]
  return feature_extractor_class(
      is_training, first_stage_features_stride,
      batch_norm_trainable, reuse_weights=reuse_weights)
Example #20
Source File: model_builder.py    From Live-feed-object-device-identification-using-Tensorflow-and-OpenCV with Apache License 2.0 5 votes vote down vote up 
def _build_faster_rcnn_feature_extractor(
    feature_extractor_config, is_training, reuse_weights=None,
    inplace_batchnorm_update=False):
  """Builds a faster_rcnn_meta_arch.FasterRCNNFeatureExtractor based on config.

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

  Returns:
    faster_rcnn_meta_arch.FasterRCNNFeatureExtractor based on config.

  Raises:
    ValueError: On invalid feature extractor type.
  """
  if inplace_batchnorm_update:
    raise ValueError('inplace batchnorm updates not supported.')
  feature_type = feature_extractor_config.type
  first_stage_features_stride = (
      feature_extractor_config.first_stage_features_stride)
  batch_norm_trainable = feature_extractor_config.batch_norm_trainable

  if feature_type not in FASTER_RCNN_FEATURE_EXTRACTOR_CLASS_MAP:
    raise ValueError('Unknown Faster R-CNN feature_extractor: {}'.format(
        feature_type))
  feature_extractor_class = FASTER_RCNN_FEATURE_EXTRACTOR_CLASS_MAP[
      feature_type]
  return feature_extractor_class(
      is_training, first_stage_features_stride,
      batch_norm_trainable, reuse_weights=reuse_weights)
Example #21
Source File: faster_rcnn_pnas_feature_extractor.py    From Live-feed-object-device-identification-using-Tensorflow-and-OpenCV with Apache License 2.0 5 votes vote down vote up 
def restore_from_classification_checkpoint_fn(
      self,
      first_stage_feature_extractor_scope,
      second_stage_feature_extractor_scope):
    """Returns a map of variables to load from a foreign checkpoint.

    Note that this overrides the default implementation in
    faster_rcnn_meta_arch.FasterRCNNFeatureExtractor which does not work for
    PNASNet checkpoints.

    Args:
      first_stage_feature_extractor_scope: A scope name for the first stage
        feature extractor.
      second_stage_feature_extractor_scope: A scope name for the second stage
        feature extractor.

    Returns:
      A dict mapping variable names (to load from a checkpoint) to variables in
      the model graph.
    """
    variables_to_restore = {}
    for variable in variables_helper.get_global_variables_safely():
      if variable.op.name.startswith(
          first_stage_feature_extractor_scope):
        var_name = variable.op.name.replace(
            first_stage_feature_extractor_scope + '/', '')
        var_name += '/ExponentialMovingAverage'
        variables_to_restore[var_name] = variable
      if variable.op.name.startswith(
          second_stage_feature_extractor_scope):
        var_name = variable.op.name.replace(
            second_stage_feature_extractor_scope + '/', '')
        var_name += '/ExponentialMovingAverage'
        variables_to_restore[var_name] = variable
    return variables_to_restore
Example #22
Source File: faster_rcnn_nas_feature_extractor.py    From Live-feed-object-device-identification-using-Tensorflow-and-OpenCV with Apache License 2.0 5 votes vote down vote up 
def restore_from_classification_checkpoint_fn(
      self,
      first_stage_feature_extractor_scope,
      second_stage_feature_extractor_scope):
    """Returns a map of variables to load from a foreign checkpoint.

    Note that this overrides the default implementation in
    faster_rcnn_meta_arch.FasterRCNNFeatureExtractor which does not work for
    NASNet-A checkpoints.

    Args:
      first_stage_feature_extractor_scope: A scope name for the first stage
        feature extractor.
      second_stage_feature_extractor_scope: A scope name for the second stage
        feature extractor.

    Returns:
      A dict mapping variable names (to load from a checkpoint) to variables in
      the model graph.
    """
    # Note that the NAS checkpoint only contains the moving average version of
    # the Variables so we need to generate an appropriate dictionary mapping.
    variables_to_restore = {}
    for variable in variables_helper.get_global_variables_safely():
      if variable.op.name.startswith(
          first_stage_feature_extractor_scope):
        var_name = variable.op.name.replace(
            first_stage_feature_extractor_scope + '/', '')
        var_name += '/ExponentialMovingAverage'
        variables_to_restore[var_name] = variable
      if variable.op.name.startswith(
          second_stage_feature_extractor_scope):
        var_name = variable.op.name.replace(
            second_stage_feature_extractor_scope + '/', '')
        var_name += '/ExponentialMovingAverage'
        variables_to_restore[var_name] = variable
    return variables_to_restore
Example #23
Source File: model_builder.py    From hands-detection with MIT License 5 votes vote down vote up 
def _build_faster_rcnn_feature_extractor(
    feature_extractor_config, is_training, reuse_weights=None):
  """Builds a faster_rcnn_meta_arch.FasterRCNNFeatureExtractor based on config.

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

  Returns:
    faster_rcnn_meta_arch.FasterRCNNFeatureExtractor based on config.

  Raises:
    ValueError: On invalid feature extractor type.
  """
  feature_type = feature_extractor_config.type
  first_stage_features_stride = (
      feature_extractor_config.first_stage_features_stride)

  if feature_type not in FASTER_RCNN_FEATURE_EXTRACTOR_CLASS_MAP:
    raise ValueError('Unknown Faster R-CNN feature_extractor: {}'.format(
        feature_type))
  feature_extractor_class = FASTER_RCNN_FEATURE_EXTRACTOR_CLASS_MAP[
      feature_type]
  return feature_extractor_class(
      is_training, first_stage_features_stride, reuse_weights)
Example #24
Source File: model_builder.py    From moveo_ros with MIT License 5 votes vote down vote up 
def _build_faster_rcnn_feature_extractor(
    feature_extractor_config, is_training, reuse_weights=None):
  """Builds a faster_rcnn_meta_arch.FasterRCNNFeatureExtractor based on config.

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

  Returns:
    faster_rcnn_meta_arch.FasterRCNNFeatureExtractor based on config.

  Raises:
    ValueError: On invalid feature extractor type.
  """
  feature_type = feature_extractor_config.type
  first_stage_features_stride = (
      feature_extractor_config.first_stage_features_stride)

  if feature_type not in FASTER_RCNN_FEATURE_EXTRACTOR_CLASS_MAP:
    raise ValueError('Unknown Faster R-CNN feature_extractor: {}'.format(
        feature_type))
  feature_extractor_class = FASTER_RCNN_FEATURE_EXTRACTOR_CLASS_MAP[
      feature_type]
  return feature_extractor_class(
      is_training, first_stage_features_stride, reuse_weights)
Example #25
Source File: model_builder.py    From BMW-TensorFlow-Training-GUI with Apache License 2.0 5 votes vote down vote up 
def _build_faster_rcnn_feature_extractor(
    feature_extractor_config, is_training, reuse_weights=None,
    inplace_batchnorm_update=False):
  """Builds a faster_rcnn_meta_arch.FasterRCNNFeatureExtractor based on config.

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

  Returns:
    faster_rcnn_meta_arch.FasterRCNNFeatureExtractor based on config.

  Raises:
    ValueError: On invalid feature extractor type.
  """
  if inplace_batchnorm_update:
    raise ValueError('inplace batchnorm updates not supported.')
  feature_type = feature_extractor_config.type
  first_stage_features_stride = (
      feature_extractor_config.first_stage_features_stride)
  batch_norm_trainable = feature_extractor_config.batch_norm_trainable

  if feature_type not in FASTER_RCNN_FEATURE_EXTRACTOR_CLASS_MAP:
    raise ValueError('Unknown Faster R-CNN feature_extractor: {}'.format(
        feature_type))
  feature_extractor_class = FASTER_RCNN_FEATURE_EXTRACTOR_CLASS_MAP[
      feature_type]
  return feature_extractor_class(
      is_training, first_stage_features_stride,
      batch_norm_trainable, reuse_weights)
Example #26
Source File: faster_rcnn_pnas_feature_extractor.py    From BMW-TensorFlow-Training-GUI with Apache License 2.0 5 votes vote down vote up 
def restore_from_classification_checkpoint_fn(
      self,
      first_stage_feature_extractor_scope,
      second_stage_feature_extractor_scope):
    """Returns a map of variables to load from a foreign checkpoint.

    Note that this overrides the default implementation in
    faster_rcnn_meta_arch.FasterRCNNFeatureExtractor which does not work for
    PNASNet checkpoints.

    Args:
      first_stage_feature_extractor_scope: A scope name for the first stage
        feature extractor.
      second_stage_feature_extractor_scope: A scope name for the second stage
        feature extractor.

    Returns:
      A dict mapping variable names (to load from a checkpoint) to variables in
      the model graph.
    """
    variables_to_restore = {}
    for variable in tf.global_variables():
      if variable.op.name.startswith(
          first_stage_feature_extractor_scope):
        var_name = variable.op.name.replace(
            first_stage_feature_extractor_scope + '/', '')
        var_name += '/ExponentialMovingAverage'
        variables_to_restore[var_name] = variable
      if variable.op.name.startswith(
          second_stage_feature_extractor_scope):
        var_name = variable.op.name.replace(
            second_stage_feature_extractor_scope + '/', '')
        var_name += '/ExponentialMovingAverage'
        variables_to_restore[var_name] = variable
    return variables_to_restore
Example #27
Source File: faster_rcnn_nas_feature_extractor.py    From BMW-TensorFlow-Training-GUI with Apache License 2.0 5 votes vote down vote up 
def restore_from_classification_checkpoint_fn(
      self,
      first_stage_feature_extractor_scope,
      second_stage_feature_extractor_scope):
    """Returns a map of variables to load from a foreign checkpoint.

    Note that this overrides the default implementation in
    faster_rcnn_meta_arch.FasterRCNNFeatureExtractor which does not work for
    NASNet-A checkpoints.

    Args:
      first_stage_feature_extractor_scope: A scope name for the first stage
        feature extractor.
      second_stage_feature_extractor_scope: A scope name for the second stage
        feature extractor.

    Returns:
      A dict mapping variable names (to load from a checkpoint) to variables in
      the model graph.
    """
    # Note that the NAS checkpoint only contains the moving average version of
    # the Variables so we need to generate an appropriate dictionary mapping.
    variables_to_restore = {}
    for variable in tf.global_variables():
      if variable.op.name.startswith(
          first_stage_feature_extractor_scope):
        var_name = variable.op.name.replace(
            first_stage_feature_extractor_scope + '/', '')
        var_name += '/ExponentialMovingAverage'
        variables_to_restore[var_name] = variable
      if variable.op.name.startswith(
          second_stage_feature_extractor_scope):
        var_name = variable.op.name.replace(
            second_stage_feature_extractor_scope + '/', '')
        var_name += '/ExponentialMovingAverage'
        variables_to_restore[var_name] = variable
    return variables_to_restore
Example #28
Source File: model_builder.py    From ros_tensorflow with Apache License 2.0 5 votes vote down vote up 
def _build_faster_rcnn_feature_extractor(
    feature_extractor_config, is_training, reuse_weights=None,
    inplace_batchnorm_update=False):
  """Builds a faster_rcnn_meta_arch.FasterRCNNFeatureExtractor based on config.

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

  Returns:
    faster_rcnn_meta_arch.FasterRCNNFeatureExtractor based on config.

  Raises:
    ValueError: On invalid feature extractor type.
  """
  if inplace_batchnorm_update:
    raise ValueError('inplace batchnorm updates not supported.')
  feature_type = feature_extractor_config.type
  first_stage_features_stride = (
      feature_extractor_config.first_stage_features_stride)
  batch_norm_trainable = feature_extractor_config.batch_norm_trainable

  if feature_type not in FASTER_RCNN_FEATURE_EXTRACTOR_CLASS_MAP:
    raise ValueError('Unknown Faster R-CNN feature_extractor: {}'.format(
        feature_type))
  feature_extractor_class = FASTER_RCNN_FEATURE_EXTRACTOR_CLASS_MAP[
      feature_type]
  return feature_extractor_class(
      is_training, first_stage_features_stride,
      batch_norm_trainable, reuse_weights)
Example #29
Source File: faster_rcnn_pnas_feature_extractor.py    From ros_tensorflow with Apache License 2.0 5 votes vote down vote up 
def restore_from_classification_checkpoint_fn(
      self,
      first_stage_feature_extractor_scope,
      second_stage_feature_extractor_scope):
    """Returns a map of variables to load from a foreign checkpoint.

    Note that this overrides the default implementation in
    faster_rcnn_meta_arch.FasterRCNNFeatureExtractor which does not work for
    PNASNet checkpoints.

    Args:
      first_stage_feature_extractor_scope: A scope name for the first stage
        feature extractor.
      second_stage_feature_extractor_scope: A scope name for the second stage
        feature extractor.

    Returns:
      A dict mapping variable names (to load from a checkpoint) to variables in
      the model graph.
    """
    variables_to_restore = {}
    for variable in tf.global_variables():
      if variable.op.name.startswith(
          first_stage_feature_extractor_scope):
        var_name = variable.op.name.replace(
            first_stage_feature_extractor_scope + '/', '')
        var_name += '/ExponentialMovingAverage'
        variables_to_restore[var_name] = variable
      if variable.op.name.startswith(
          second_stage_feature_extractor_scope):
        var_name = variable.op.name.replace(
            second_stage_feature_extractor_scope + '/', '')
        var_name += '/ExponentialMovingAverage'
        variables_to_restore[var_name] = variable
    return variables_to_restore
Example #30
Source File: faster_rcnn_nas_feature_extractor.py    From ros_tensorflow with Apache License 2.0 5 votes vote down vote up 
def restore_from_classification_checkpoint_fn(
      self,
      first_stage_feature_extractor_scope,
      second_stage_feature_extractor_scope):
    """Returns a map of variables to load from a foreign checkpoint.

    Note that this overrides the default implementation in
    faster_rcnn_meta_arch.FasterRCNNFeatureExtractor which does not work for
    NASNet-A checkpoints.

    Args:
      first_stage_feature_extractor_scope: A scope name for the first stage
        feature extractor.
      second_stage_feature_extractor_scope: A scope name for the second stage
        feature extractor.

    Returns:
      A dict mapping variable names (to load from a checkpoint) to variables in
      the model graph.
    """
    # Note that the NAS checkpoint only contains the moving average version of
    # the Variables so we need to generate an appropriate dictionary mapping.
    variables_to_restore = {}
    for variable in tf.global_variables():
      if variable.op.name.startswith(
          first_stage_feature_extractor_scope):
        var_name = variable.op.name.replace(
            first_stage_feature_extractor_scope + '/', '')
        var_name += '/ExponentialMovingAverage'
        variables_to_restore[var_name] = variable
      if variable.op.name.startswith(
          second_stage_feature_extractor_scope):
        var_name = variable.op.name.replace(
            second_stage_feature_extractor_scope + '/', '')
        var_name += '/ExponentialMovingAverage'
        variables_to_restore[var_name] = variable
    return variables_to_restore