def _build_tail(self, inputs, is_training=False):
if not self._use_tail:
return inputs
if self._architecture == "resnet_v1_101":
train_batch_norm = is_training and self._config.get("train_batch_norm")
with self._enter_variable_scope():
weight_decay = self._config.get("arg_scope", {}).get("weight_decay", 0)
with tf.variable_scope(self._architecture, reuse=True):
resnet_arg_scope = resnet_utils.resnet_arg_scope(
batch_norm_epsilon=1e-5, batch_norm_scale=True, weight_decay=weight_decay
)
with slim.arg_scope(resnet_arg_scope):
with slim.arg_scope([slim.batch_norm], is_training=train_batch_norm):
blocks = [
resnet_utils.Block(
"block4",
resnet_v1.bottleneck,
[{"depth": 2048, "depth_bottleneck": 512, "stride": 1}] * 3,
)
]
proposal_classifier_features = resnet_utils.stack_blocks_dense(inputs, blocks)
else:
proposal_classifier_features = inputs
return proposal_classifier_features
def _resnet_plain(self, inputs, blocks, output_stride=None, scope=None):
"""A plain ResNet without extra layers before or after the ResNet blocks."""
with variable_scope.variable_scope(scope, values=[inputs]):
with arg_scope([layers.conv2d], outputs_collections='end_points'):
net = resnet_utils.stack_blocks_dense(inputs, blocks,
output_stride)
end_points = utils.convert_collection_to_dict('end_points')
return net, end_points
def testAtrousValuesBottleneck(self):
"""Verify the values of dense feature extraction by atrous convolution.
Make sure that dense feature extraction by stack_blocks_dense() followed by
subsampling gives identical results to feature extraction at the nominal
network output stride using the simple self._stack_blocks_nondense() above.
"""
block = resnet_v1.resnet_v1_block
blocks = [
block('block1', base_depth=1, num_units=2, stride=2),
block('block2', base_depth=2, num_units=2, stride=2),
block('block3', base_depth=4, num_units=2, stride=2),
block('block4', base_depth=8, num_units=2, stride=1),
]
nominal_stride = 8
# Test both odd and even input dimensions.
height = 30
width = 31
with arg_scope(resnet_utils.resnet_arg_scope()):
with arg_scope([layers.batch_norm], is_training=False):
for output_stride in [1, 2, 4, 8, None]:
with ops.Graph().as_default():
with self.cached_session() as sess:
random_seed.set_random_seed(0)
inputs = create_input(1, height, width, 3)
# Dense feature extraction followed by subsampling.
output = resnet_utils.stack_blocks_dense(
inputs, blocks, output_stride)
if output_stride is None:
factor = 1
else:
factor = nominal_stride // output_stride
output = resnet_utils.subsample(output, factor)
# Make the two networks use the same weights.
variable_scope.get_variable_scope(
).reuse_variables()
# Feature extraction at the nominal network rate.
expected = self._stack_blocks_nondense(
inputs, blocks)
sess.run(variables.global_variables_initializer())
output, expected = sess.run([output, expected])
self.assertAllClose(output,
expected,
atol=1e-4,
rtol=1e-4)
def _build_tail(self, inputs, is_training=False):
if not self._use_tail:
return inputs
if self._architecture == 'resnet_v1_101':
train_batch_norm = (
is_training and self._config.get('train_batch_norm')
)
with self._enter_variable_scope():
weight_decay = (
self._config.get('arg_scope', {}).get('weight_decay', 0)
)
with tf.variable_scope(self._architecture, reuse=True):
resnet_arg_scope = resnet_utils.resnet_arg_scope(
batch_norm_epsilon=1e-5,
batch_norm_scale=True,
weight_decay=weight_decay
)
with slim.arg_scope(resnet_arg_scope):
with slim.arg_scope(
[slim.batch_norm], is_training=train_batch_norm
):
blocks = [
resnet_utils.Block(
'block4',
resnet_v1.bottleneck,
[{
'depth': 2048,
'depth_bottleneck': 512,
'stride': 1
}] * 3
)
]
proposal_classifier_features = (
resnet_utils.stack_blocks_dense(inputs, blocks)
)
else:
proposal_classifier_features = inputs
return proposal_classifier_features
def resnet_v2(inputs,
blocks,
num_classes=None,
is_training=True,
global_pool=True,
output_stride=None,
include_root_block=True,
reuse=None,
scope=None):
"""Generator for v2 (preactivation) ResNet models.
This function generates a family of ResNet v2 models. See the resnet_v2_*()
methods for specific model instantiations, obtained by selecting different
block instantiations that produce ResNets of various depths.
Training for image classification on Imagenet is usually done with [224, 224]
inputs, resulting in [7, 7] feature maps at the output of the last ResNet
block for the ResNets defined in [1] that have nominal stride equal to 32.
However, for dense prediction tasks we advise that one uses inputs with
spatial dimensions that are multiples of 32 plus 1, e.g., [321, 321]. In
this case the feature maps at the ResNet output will have spatial shape
[(height - 1) / output_stride + 1, (width - 1) / output_stride + 1]
and corners exactly aligned with the input image corners, which greatly
facilitates alignment of the features to the image. Using as input [225, 225]
images results in [8, 8] feature maps at the output of the last ResNet block.
For dense prediction tasks, the ResNet needs to run in fully-convolutional
(FCN) mode and global_pool needs to be set to False. The ResNets in [1, 2] all
have nominal stride equal to 32 and a good choice in FCN mode is to use
output_stride=16 in order to increase the density of the computed features at
small computational and memory overhead, cf. http://arxiv.org/abs/1606.00915.
Args:
inputs: A tensor of size [batch, height_in, width_in, channels].
blocks: A list of length equal to the number of ResNet blocks. Each element
is a resnet_utils.Block object describing the units in the block.
num_classes: Number of predicted classes for classification tasks. If None
we return the features before the logit layer.
is_training: whether batch_norm layers are in training mode.
global_pool: If True, we perform global average pooling before computing the
logits. Set to True for image classification, False for dense prediction.
output_stride: If None, then the output will be computed at the nominal
network stride. If output_stride is not None, it specifies the requested
ratio of input to output spatial resolution.
include_root_block: If True, include the initial convolution followed by
max-pooling, if False excludes it. If excluded, `inputs` should be the
results of an activation-less convolution.
reuse: whether or not the network and its variables should be reused. To be
able to reuse 'scope' must be given.
scope: Optional variable_scope.
Returns:
net: A rank-4 tensor of size [batch, height_out, width_out, channels_out].
If global_pool is False, then height_out and width_out are reduced by a
factor of output_stride compared to the respective height_in and width_in,
else both height_out and width_out equal one. If num_classes is None, then
net is the output of the last ResNet block, potentially after global
average pooling. If num_classes is not None, net contains the pre-softmax
activations.
end_points: A dictionary from components of the network to the corresponding
activation.
Raises:
ValueError: If the target output_stride is not valid.
"""
with variable_scope.variable_scope(scope,
'resnet_v2', [inputs],
reuse=reuse) as sc:
end_points_collection = sc.original_name_scope + '_end_points'
with arg_scope(
[layers_lib.conv2d, bottleneck, resnet_utils.stack_blocks_dense],
outputs_collections=end_points_collection):
with arg_scope([layers.batch_norm], is_training=is_training):
net = inputs
if include_root_block:
if output_stride is not None:
if output_stride % 4 != 0:
raise ValueError(
'The output_stride needs to be a multiple of 4.'
)
output_stride /= 4
# We do not include batch normalization or activation functions in
# conv1 because the first ResNet unit will perform these. Cf.
# Appendix of [2].
with arg_scope([layers_lib.conv2d],
activation_fn=None,
normalizer_fn=None):
net = resnet_utils.conv2d_same(net,
64,
7,
stride=2,
scope='conv1')
net = layers.max_pool2d(net, [3, 3],
stride=2,
scope='pool1')
net = resnet_utils.stack_blocks_dense(net, blocks,
output_stride)
# This is needed because the pre-activation variant does not have batch
# normalization or activation functions in the residual unit output. See
# Appendix of [2].
net = layers.batch_norm(net,
activation_fn=nn_ops.relu,
scope='postnorm')
if global_pool:
# Global average pooling.
net = math_ops.reduce_mean(net, [1, 2],
name='pool5',
keepdims=True)
if num_classes is not None:
net = layers_lib.conv2d(net,
num_classes, [1, 1],
activation_fn=None,
normalizer_fn=None,
scope='logits')
# Convert end_points_collection into a dictionary of end_points.
end_points = utils.convert_collection_to_dict(
end_points_collection)
if num_classes is not None:
end_points['predictions'] = layers.softmax(
net, scope='predictions')
return net, end_points