def test_prediction_size_depthwise_false(self):
conv_hyperparams = self._build_conv_hyperparams()
class_prediction_head = keras_class_head.ConvolutionalClassHead(
is_training=True,
num_class_slots=20,
use_dropout=True,
dropout_keep_prob=0.5,
kernel_size=3,
conv_hyperparams=conv_hyperparams,
freeze_batchnorm=False,
num_predictions_per_location=1,
use_depthwise=False)
image_feature = tf.random_uniform(
[64, 17, 19, 1024], minval=-10.0, maxval=10.0, dtype=tf.float32)
class_predictions = class_prediction_head(image_feature,)
self.assertAllEqual([64, 323, 20],
class_predictions.get_shape().as_list())
def build_convolutional_keras_box_predictor(is_training,
num_classes,
conv_hyperparams,
freeze_batchnorm,
inplace_batchnorm_update,
num_predictions_per_location_list,
min_depth,
max_depth,
num_layers_before_predictor,
use_dropout,
dropout_keep_prob,
kernel_size,
box_code_size,
add_background_class=True,
class_prediction_bias_init=0.0,
use_depthwise=False,
name='BoxPredictor'):
"""Builds the Keras ConvolutionalBoxPredictor from the arguments.
Args:
is_training: Indicates whether the BoxPredictor is in training mode.
num_classes: number of classes. Note that num_classes *does not*
include the background category, so if groundtruth labels take values
in {0, 1, .., K-1}, num_classes=K (and not K+1, even though the
assigned classification targets can range from {0,... K}).
conv_hyperparams: A `hyperparams_builder.KerasLayerHyperparams` object
containing hyperparameters for convolution ops.
freeze_batchnorm: Whether to freeze batch norm parameters during
training or not. When training with a small batch size (e.g. 1), it is
desirable to freeze batch norm update and use pretrained batch norm
params.
inplace_batchnorm_update: Whether to update batch norm moving average
values inplace. When this is false train op must add a control
dependency on tf.graphkeys.UPDATE_OPS collection in order to update
batch norm statistics.
num_predictions_per_location_list: A list of integers representing the
number of box predictions to be made per spatial location for each
feature map.
min_depth: Minimum feature depth prior to predicting box encodings
and class predictions.
max_depth: Maximum feature depth prior to predicting box encodings
and class predictions. If max_depth is set to 0, no additional
feature map will be inserted before location and class predictions.
num_layers_before_predictor: Number of the additional conv layers before
the predictor.
use_dropout: Option to use dropout or not. Note that a single dropout
op is applied here prior to both box and class predictions, which stands
in contrast to the ConvolutionalBoxPredictor below.
dropout_keep_prob: Keep probability for dropout.
This is only used if use_dropout is True.
kernel_size: Size of final convolution kernel. If the
spatial resolution of the feature map is smaller than the kernel size,
then the kernel size is automatically set to be
min(feature_width, feature_height).
box_code_size: Size of encoding for each box.
add_background_class: Whether to add an implicit background class.
class_prediction_bias_init: constant value to initialize bias of the last
conv2d layer before class prediction.
use_depthwise: Whether to use depthwise convolutions for prediction
steps. Default is False.
name: A string name scope to assign to the box predictor. If `None`, Keras
will auto-generate one from the class name.
Returns:
A Keras ConvolutionalBoxPredictor class.
"""
box_prediction_heads = []
class_prediction_heads = []
other_heads = {}
for stack_index, num_predictions_per_location in enumerate(
num_predictions_per_location_list):
box_prediction_heads.append(
keras_box_head.ConvolutionalBoxHead(
is_training=is_training,
box_code_size=box_code_size,
kernel_size=kernel_size,
conv_hyperparams=conv_hyperparams,
freeze_batchnorm=freeze_batchnorm,
num_predictions_per_location=num_predictions_per_location,
use_depthwise=use_depthwise,
name='ConvolutionalBoxHead_%d' % stack_index))
class_prediction_heads.append(
keras_class_head.ConvolutionalClassHead(
is_training=is_training,
num_class_slots=(num_classes +
1 if add_background_class else num_classes),
use_dropout=use_dropout,
dropout_keep_prob=dropout_keep_prob,
kernel_size=kernel_size,
conv_hyperparams=conv_hyperparams,
freeze_batchnorm=freeze_batchnorm,
num_predictions_per_location=num_predictions_per_location,
class_prediction_bias_init=class_prediction_bias_init,
use_depthwise=use_depthwise,
name='ConvolutionalClassHead_%d' % stack_index))
return convolutional_keras_box_predictor.ConvolutionalBoxPredictor(
is_training=is_training,
num_classes=num_classes,
box_prediction_heads=box_prediction_heads,
class_prediction_heads=class_prediction_heads,
other_heads=other_heads,
conv_hyperparams=conv_hyperparams,
num_layers_before_predictor=num_layers_before_predictor,
min_depth=min_depth,
max_depth=max_depth,
freeze_batchnorm=freeze_batchnorm,
inplace_batchnorm_update=inplace_batchnorm_update,
name=name)
def build_convolutional_keras_box_predictor(is_training,
num_classes,
conv_hyperparams,
freeze_batchnorm,
inplace_batchnorm_update,
num_predictions_per_location_list,
min_depth,
max_depth,
num_layers_before_predictor,
use_dropout,
dropout_keep_prob,
kernel_size,
box_code_size,
class_prediction_bias_init=0.0,
use_depthwise=False,
mask_head_config=None,
name='BoxPredictor'):
"""Builds the ConvolutionalBoxPredictor from the arguments.
Args:
is_training: Indicates whether the BoxPredictor is in training mode.
num_classes: Number of classes.
conv_hyperparams: A `hyperparams_builder.KerasLayerHyperparams` object
containing hyperparameters for convolution ops.
freeze_batchnorm: Whether to freeze batch norm parameters during
training or not. When training with a small batch size (e.g. 1), it is
desirable to freeze batch norm update and use pretrained batch norm
params.
inplace_batchnorm_update: Whether to update batch norm moving average
values inplace. When this is false train op must add a control
dependency on tf.graphkeys.UPDATE_OPS collection in order to update
batch norm statistics.
num_predictions_per_location_list: A list of integers representing the
number of box predictions to be made per spatial location for each
feature map.
min_depth: Minimum feature depth prior to predicting box encodings
and class predictions.
max_depth: Maximum feature depth prior to predicting box encodings
and class predictions. If max_depth is set to 0, no additional
feature map will be inserted before location and class predictions.
num_layers_before_predictor: Number of the additional conv layers before
the predictor.
use_dropout: Option to use dropout or not. Note that a single dropout
op is applied here prior to both box and class predictions, which stands
in contrast to the ConvolutionalBoxPredictor below.
dropout_keep_prob: Keep probability for dropout.
This is only used if use_dropout is True.
kernel_size: Size of final convolution kernel. If the
spatial resolution of the feature map is smaller than the kernel size,
then the kernel size is automatically set to be
min(feature_width, feature_height).
box_code_size: Size of encoding for each box.
class_prediction_bias_init: constant value to initialize bias of the last
conv2d layer before class prediction.
use_depthwise: Whether to use depthwise convolutions for prediction
steps. Default is False.
mask_head_config: An optional MaskHead object containing configs for mask
head construction.
name: A string name scope to assign to the box predictor. If `None`, Keras
will auto-generate one from the class name.
Returns:
A ConvolutionalBoxPredictor class.
"""
box_prediction_heads = []
class_prediction_heads = []
mask_prediction_heads = []
other_heads = {}
if mask_head_config is not None:
other_heads[convolutional_box_predictor.MASK_PREDICTIONS] = \
mask_prediction_heads
for stack_index, num_predictions_per_location in enumerate(
num_predictions_per_location_list):
box_prediction_heads.append(
keras_box_head.ConvolutionalBoxHead(
is_training=is_training,
box_code_size=box_code_size,
kernel_size=kernel_size,
conv_hyperparams=conv_hyperparams,
freeze_batchnorm=freeze_batchnorm,
num_predictions_per_location=num_predictions_per_location,
use_depthwise=use_depthwise,
name='ConvolutionalBoxHead_%d' % stack_index))
class_prediction_heads.append(
keras_class_head.ConvolutionalClassHead(
is_training=is_training,
num_classes=num_classes,
use_dropout=use_dropout,
dropout_keep_prob=dropout_keep_prob,
kernel_size=kernel_size,
conv_hyperparams=conv_hyperparams,
freeze_batchnorm=freeze_batchnorm,
num_predictions_per_location=num_predictions_per_location,
class_prediction_bias_init=class_prediction_bias_init,
use_depthwise=use_depthwise,
name='ConvolutionalClassHead_%d' % stack_index))
if mask_head_config is not None:
if not mask_head_config.masks_are_class_agnostic:
logging.warning(
'Note that class specific mask prediction for SSD '
'models is memory consuming.')
mask_prediction_heads.append(
keras_mask_head.ConvolutionalMaskHead(
is_training=is_training,
num_classes=num_classes,
use_dropout=use_dropout,
dropout_keep_prob=dropout_keep_prob,
kernel_size=kernel_size,
conv_hyperparams=conv_hyperparams,
freeze_batchnorm=freeze_batchnorm,
num_predictions_per_location=num_predictions_per_location,
use_depthwise=use_depthwise,
mask_height=mask_head_config.mask_height,
mask_width=mask_head_config.mask_width,
masks_are_class_agnostic=mask_head_config.
masks_are_class_agnostic,
name='ConvolutionalMaskHead_%d' % stack_index))
return convolutional_keras_box_predictor.ConvolutionalBoxPredictor(
is_training=is_training,
num_classes=num_classes,
box_prediction_heads=box_prediction_heads,
class_prediction_heads=class_prediction_heads,
other_heads=other_heads,
conv_hyperparams=conv_hyperparams,
num_layers_before_predictor=num_layers_before_predictor,
min_depth=min_depth,
max_depth=max_depth,
freeze_batchnorm=freeze_batchnorm,
inplace_batchnorm_update=inplace_batchnorm_update,
name=name)
def build_convolutional_keras_box_predictor(is_training,
num_classes,
conv_hyperparams,
freeze_batchnorm,
inplace_batchnorm_update,
num_predictions_per_location_list,
min_depth,
max_depth,
num_layers_before_predictor,
use_dropout,
dropout_keep_prob,
kernel_size,
box_code_size,
add_background_class=True,
class_prediction_bias_init=0.0,
use_depthwise=False,
box_encodings_clip_range=None,
name='BoxPredictor'):
box_prediction_heads = []
class_prediction_heads = []
other_heads = {}
for stack_index, num_predictions_per_location in enumerate(
num_predictions_per_location_list):
box_prediction_heads.append(
keras_box_head.ConvolutionalBoxHead(
is_training=is_training,
box_code_size=box_code_size,
kernel_size=kernel_size,
conv_hyperparams=conv_hyperparams,
freeze_batchnorm=freeze_batchnorm,
num_predictions_per_location=num_predictions_per_location,
use_depthwise=use_depthwise,
box_encodings_clip_range=box_encodings_clip_range,
name='ConvolutionalBoxHead_%d' % stack_index))
class_prediction_heads.append(
keras_class_head.ConvolutionalClassHead(
is_training=is_training,
num_class_slots=(
num_classes + 1 if add_background_class else num_classes),
use_dropout=use_dropout,
dropout_keep_prob=dropout_keep_prob,
kernel_size=kernel_size,
conv_hyperparams=conv_hyperparams,
freeze_batchnorm=freeze_batchnorm,
num_predictions_per_location=num_predictions_per_location,
class_prediction_bias_init=class_prediction_bias_init,
use_depthwise=use_depthwise,
name='ConvolutionalClassHead_%d' % stack_index))
return convolutional_keras_box_predictor.ConvolutionalBoxPredictor(
is_training=is_training,
num_classes=num_classes,
box_prediction_heads=box_prediction_heads,
class_prediction_heads=class_prediction_heads,
other_heads=other_heads,
conv_hyperparams=conv_hyperparams,
num_layers_before_predictor=num_layers_before_predictor,
min_depth=min_depth,
max_depth=max_depth,
freeze_batchnorm=freeze_batchnorm,
inplace_batchnorm_update=inplace_batchnorm_update,
name=name)
