def block_group(inputs,
filters,
strides,
use_projection,
block_fn,
block_repeats,
batch_norm_relu=nn_ops.BatchNormRelu(),
dropblock=nn_ops.Dropblock(),
drop_connect_rate=None,
data_format='channels_last',
name=None,
is_training=False):
"""Builds one group of blocks.
Args:
inputs: a `Tensor` of size `[batch, channels, height, width]`.
filters: an `int` number of filters for the first two convolutions.
strides: an `int` block stride. If greater than 1, this block will
ultimately downsample the input.
use_projection: a `bool` for whether this block should use a projection
shortcut (versus the default identity shortcut). This is usually `True`
for the first block of a block group, which may change the number of
filters and the resolution.
block_fn: the `function` for the block to use within the model
block_repeats: an `int` number of blocks to repeat in the group.
batch_norm_relu: an operation that is added after convolutions, including a
batch norm layer and an optional relu activation.
dropblock: a drop block layer that is added after convluations. Note that
the default implementation does not apply any drop block.
drop_connect_rate: a 'float' number that specifies the drop connection rate
of the block. Note that the default `None` means no drop connection is
applied.
data_format: a `str` that specifies the data format.
name: a `str` name for the Tensor output of the block layer.
is_training: a `bool` if True, the model is in training mode.
Returns:
The output `Tensor` of the block layer.
"""
# Only the first block per block_group uses projection shortcut and strides.
inputs = block_fn(inputs,
filters,
strides,
use_projection=use_projection,
batch_norm_relu=batch_norm_relu,
dropblock=dropblock,
drop_connect_rate=drop_connect_rate,
data_format=data_format,
is_training=is_training)
for _ in range(1, block_repeats):
inputs = block_fn(inputs,
filters,
1,
use_projection=False,
batch_norm_relu=batch_norm_relu,
dropblock=dropblock,
drop_connect_rate=drop_connect_rate,
data_format=data_format,
is_training=is_training)
return tf.identity(inputs, name)
def __init__(self,
resnet_depth,
dropblock=nn_ops.Dropblock(),
batch_norm_relu=nn_ops.BatchNormRelu(),
data_format='channels_last'):
"""ResNet initialization function.
Args:
resnet_depth: `int` depth of ResNet backbone model.
dropblock: a dropblock layer.
batch_norm_relu: an operation that includes a batch normalization layer
followed by a relu layer(optional).
data_format: `str` either "channels_first" for `[batch, channels, height,
width]` or "channels_last for `[batch, height, width, channels]`.
"""
self._resnet_depth = resnet_depth
self._dropblock = dropblock
self._batch_norm_relu = batch_norm_relu
self._data_format = data_format
model_params = {
18: {
'block': self.residual_block,
'layers': [2, 2, 2, 2]
},
34: {
'block': self.residual_block,
'layers': [3, 4, 6, 3]
},
50: {
'block': self.bottleneck_block,
'layers': [3, 4, 6, 3]
},
101: {
'block': self.bottleneck_block,
'layers': [3, 4, 23, 3]
},
152: {
'block': self.bottleneck_block,
'layers': [3, 8, 36, 3]
},
200: {
'block': self.bottleneck_block,
'layers': [3, 24, 36, 3]
}
}
if resnet_depth not in model_params:
valid_resnet_depths = ', '.join(
[str(depth) for depth in sorted(model_params.keys())])
raise ValueError(
'The resnet_depth should be in [%s]. Not a valid resnet_depth:'
% (valid_resnet_depths), self._resnet_depth)
params = model_params[resnet_depth]
self._resnet_fn = self.resnet_v1_generator(params['block'],
params['layers'])
def __init__(self,
min_level=3,
max_level=7,
block_specs=build_block_specs(),
endpoints_num_filters=48,
use_native_resize_op=False,
se_ratio=0.2,
block_repeats=1,
filter_size_scale=1.0,
activation='swish',
batch_norm_activation=nn_ops.BatchNormActivation(
activation='swish'),
init_drop_connect_rate=None,
data_format='channels_last'):
"""SpineNetMBConv initialization function.
Args:
min_level: `int` minimum level in SpineNet endpoints.
max_level: `int` maximum level in SpineNet endpoints.
block_specs: a list of BlockSpec objects that specifies the SpineNet
network topology. By default, the previously discovered architecture is
used.
endpoints_num_filters: `int` feature dimension applied to endpoints before
sharing conv layers in head.
use_native_resize_op: Whether to use native
tf.image.nearest_neighbor_resize or the broadcast implmentation to do
upsampling.
se_ratio: squeeze and excitation ratio for MBConv blocks.
block_repeats: `int` number of repeats per block.
filter_size_scale: `float` a scaling factor to uniformaly scale feature
dimension in SpineNet.
activation: the activation function after cross-scale feature fusion.
Support 'relu' and 'swish'.
batch_norm_activation: An operation that includes a batch normalization
layer followed by an optional activation layer.
init_drop_connect_rate: `float` initial drop connect rate.
data_format: An optional string from: "channels_last", "channels_first".
Defaults to "channels_last".
"""
self._min_level = min_level
self._max_level = max_level
self._block_specs = block_specs
self._endpoints_num_filters = endpoints_num_filters
self._use_native_resize_op = use_native_resize_op
self._se_ratio = se_ratio
self._block_repeats = block_repeats
self._filter_size_scale = filter_size_scale
if activation == 'relu':
self._activation = tf.nn.relu
elif activation == 'swish':
self._activation = tf.nn.swish
else:
raise ValueError(
'Activation {} not implemented.'.format(activation))
self._batch_norm_activation = batch_norm_activation
self._init_dc_rate = init_drop_connect_rate
self._data_format = data_format
self._dropblock = nn_ops.Dropblock()
def block_group(inputs,
filters,
strides,
block_fn_cand,
block_repeats,
activation=tf.nn.swish,
batch_norm_activation=nn_ops.BatchNormActivation(),
dropblock=nn_ops.Dropblock(),
drop_connect_rate=None,
data_format='channels_last',
name=None,
is_training=False):
"""Creates one group of blocks for SpineNet."""
block_fn_candidates = {
'bottleneck': nn_blocks.bottleneck_block,
'residual': nn_blocks.residual_block,
}
if block_fn_cand not in block_fn_candidates:
raise ValueError('Block function {} not implemented.'.format(block_fn_cand))
block_fn = block_fn_candidates[block_fn_cand]
_, _, _, num_filters = inputs.get_shape().as_list()
if block_fn_cand == 'bottleneck':
use_projection = not (num_filters == (filters * 4) and strides == 1)
else:
use_projection = not (num_filters == filters and strides == 1)
# Only the first block per block_group uses projection shortcut and strides.
inputs = block_fn(
inputs,
filters,
strides,
use_projection=use_projection,
activation=activation,
batch_norm_activation=batch_norm_activation,
dropblock=dropblock,
drop_connect_rate=drop_connect_rate,
data_format=data_format,
is_training=is_training)
for _ in range(1, block_repeats):
inputs = block_fn(
inputs,
filters,
1,
use_projection=False,
activation=activation,
batch_norm_activation=batch_norm_activation,
dropblock=dropblock,
drop_connect_rate=drop_connect_rate,
data_format=data_format,
is_training=is_training)
return tf.identity(inputs, name)
def block_group(inputs,
filters,
strides,
block_fn,
block_repeats,
conv2d_op=None,
activation=tf.nn.swish,
batch_norm_activation=nn_ops.BatchNormActivation(),
dropblock=nn_ops.Dropblock(),
drop_connect_rate=None,
data_format='channels_last',
name=None,
is_training=False):
"""Creates one group of blocks for NAS-FPN."""
if block_fn == 'conv':
inputs = conv2d_op(inputs,
filters=filters,
kernel_size=(3, 3),
padding='same',
data_format=data_format,
name='conv')
inputs = batch_norm_activation(inputs,
is_training=is_training,
relu=False,
name='bn')
inputs = dropblock(inputs, is_training=is_training)
return inputs
if block_fn != 'bottleneck':
raise ValueError('Block function {} not implemented.'.format(block_fn))
_, _, _, num_filters = inputs.get_shape().as_list()
block_fn = nn_blocks.bottleneck_block
use_projection = not (num_filters == (filters * 4) and strides == 1)
return resnet.block_group(inputs=inputs,
filters=filters,
strides=strides,
use_projection=use_projection,
block_fn=block_fn,
block_repeats=block_repeats,
activation=activation,
batch_norm_activation=batch_norm_activation,
dropblock=dropblock,
drop_connect_rate=drop_connect_rate,
data_format=data_format,
name=name,
is_training=is_training)
def __init__(self,
min_level=3,
max_level=7,
fpn_feat_dims=256,
num_repeats=7,
use_separable_conv=False,
dropblock=nn_ops.Dropblock(),
batch_norm_relu=nn_ops.BatchNormRelu()):
"""NAS-FPN initialization function.
Args:
min_level: `int` minimum level in NAS-FPN output feature maps.
max_level: `int` maximum level in NAS-FPN output feature maps.
fpn_feat_dims: `int` number of filters in FPN layers.
num_repeats: number of repeats for feature pyramid network.
use_separable_conv: `bool`, if True use separable convolution for
convolution in NAS-FPN layers.
dropblock: a Dropblock layer.
batch_norm_relu: an operation that includes a batch normalization layer
followed by a relu layer(optional).
"""
self._min_level = min_level
self._max_level = max_level
if min_level == 3 and max_level == 7:
model_config = [
3, 1, 1, 3,
3, 0, 1, 5,
4, 0, 0, 6, # Output to level 3.
3, 0, 6, 7, # Output to level 4.
2, 1, 7, 8, # Output to level 5.
0, 1, 6, 9, # Output to level 7.
1, 1, 9, 10] # Output to level 6.
else:
raise ValueError('The NAS-FPN with min level {} and max level {} '
'is not supported.'.format(min_level, max_level))
self._config = Config(model_config, self._min_level, self._max_level)
self._num_repeats = num_repeats
self._fpn_feat_dims = fpn_feat_dims
self._use_separable_conv = use_separable_conv
self._dropblock = dropblock
self._batch_norm_relu = batch_norm_relu
self._resample_feature_map = functools.partial(
resample_feature_map,
target_feat_dims=fpn_feat_dims, batch_norm_relu=batch_norm_relu)
def block_group(inputs,
in_filters,
out_filters,
strides,
expand_ratio,
block_repeats,
se_ratio=0.2,
batch_norm_activation=nn_ops.BatchNormActivation(),
dropblock=nn_ops.Dropblock(),
drop_connect_rate=None,
data_format='channels_last',
name=None,
is_training=False):
"""Creates one group of blocks for Mobile SpineNet."""
# Apply strides only to the first block in block_group.
inputs = nn_blocks.mbconv_block(
inputs,
in_filters,
out_filters,
expand_ratio,
strides,
se_ratio=se_ratio,
batch_norm_activation=batch_norm_activation,
dropblock=dropblock,
drop_connect_rate=drop_connect_rate,
data_format=data_format,
is_training=is_training)
for _ in range(1, block_repeats):
inputs = nn_blocks.mbconv_block(
inputs,
out_filters,
out_filters,
expand_ratio,
1, # strides
se_ratio=se_ratio,
batch_norm_activation=batch_norm_activation,
dropblock=dropblock,
drop_connect_rate=drop_connect_rate,
data_format=data_format,
is_training=is_training)
return tf.identity(inputs, name)
def __init__(self,
min_level=3,
max_level=7,
block_specs=build_block_specs(),
endpoints_num_filters=256,
resample_alpha=0.5,
use_native_resize_op=False,
block_repeats=1,
filter_size_scale=1.0,
activation='swish',
batch_norm_activation=nn_ops.BatchNormActivation(
activation='swish'),
init_drop_connect_rate=None,
data_format='channels_last'):
"""SpineNet initialization function.
Args:
min_level: an `int` representing the minimum level in SpineNet endpoints.
max_level: an `int` representing the maximum level in SpineNet endpoints.
block_specs: a list of BlockSpec objects that specifies the SpineNet
network topology. By default, the previously discovered architecture is
used.
endpoints_num_filters: an `int` representing the final feature dimension
of endpoints before the shared conv layers in head.
resample_alpha: a `float` representing the scaling factor to scale feature
dimension before resolution resampling.
use_native_resize_op: Whether to use native
tf.image.nearest_neighbor_resize or the broadcast implmentation to do
upsampling.
block_repeats: an `int` representing the number of repeats per block
group.
filter_size_scale: a `float` representing the scaling factor to uniformaly
scale feature dimension in SpineNet.
activation: activation function. Support 'relu' and 'swish'.
batch_norm_activation: an operation that includes a batch normalization
layer followed by an optional activation layer.
init_drop_connect_rate: a 'float' number that specifies the initial drop
connection rate. Note that the default `None` means no drop connection
is applied.
data_format: An optional string from: "channels_last", "channels_first".
Defaults to "channels_last".
"""
self._min_level = min_level
self._max_level = max_level
self._block_specs = block_specs
self._endpoints_num_filters = endpoints_num_filters
self._use_native_resize_op = use_native_resize_op
self._resample_alpha = resample_alpha
self._block_repeats = block_repeats
self._filter_size_scale = filter_size_scale
if activation == 'relu':
self._activation = tf.nn.relu
elif activation == 'swish':
self._activation = tf.nn.swish
else:
raise ValueError(
'Activation {} not implemented.'.format(activation))
self._batch_norm_activation = batch_norm_activation
self._init_drop_connect_rate = init_drop_connect_rate
self._data_format = data_format
self._dropblock = nn_ops.Dropblock(
) # Hard-code it to not use DropBlock.
self._init_block_fn = 'bottleneck'
self._num_init_blocks = 2
def dropblock_generator(params):
return nn_ops.Dropblock(
dropblock_keep_prob=params.dropblock_keep_prob,
dropblock_size=params.dropblock_size)
def bottleneck_block(inputs,
filters,
strides,
use_projection,
activation=tf.nn.relu,
batch_norm_relu=nn_ops.BatchNormRelu(),
dropblock=nn_ops.Dropblock(),
drop_connect_rate=None,
data_format='channels_last',
is_training=False):
"""The bottleneck block with BN and DropBlock after convolutions.
Args:
inputs: a `Tensor` of size `[batch, channels, height, width]`.
filters: a `int` number of filters for the first two convolutions. Note that
the third and final convolution will use 4 times as many filters.
strides: an `int` block stride. If greater than 1, this block will
ultimately downsample the input.
use_projection: a `bool` for whether this block should use a projection
shortcut (versus the default identity shortcut). This is usually `True`
for the first block of a block group, which may change the number of
filters and the resolution.
activation: activation function. Support 'relu' and 'swish'.
batch_norm_relu: an operation that is added after convolutions, including a
batch norm layer and an optional relu activation.
dropblock: a drop block layer that is added after convluations. Note that
the default implementation does not apply any drop block.
drop_connect_rate: a 'float' number that specifies the drop connection rate
of the block. Note that the default `None` means no drop connection is
applied.
data_format: a `str` that specifies the data format.
is_training: a `bool` if True, the model is in training mode.
Returns:
The output `Tensor` of the block.
"""
logging.info('-----> Building bottleneck block.')
shortcut = inputs
if use_projection:
out_filters = 4 * filters
shortcut = nn_ops.conv2d_fixed_padding(inputs=inputs,
filters=out_filters,
kernel_size=1,
strides=strides,
data_format=data_format)
shortcut = batch_norm_relu(shortcut,
relu=False,
is_training=is_training)
shortcut = dropblock(shortcut, is_training=is_training)
inputs = nn_ops.conv2d_fixed_padding(inputs=inputs,
filters=filters,
kernel_size=1,
strides=1,
data_format=data_format)
inputs = batch_norm_relu(inputs, is_training=is_training)
inputs = dropblock(inputs, is_training=is_training)
inputs = nn_ops.conv2d_fixed_padding(inputs=inputs,
filters=filters,
kernel_size=3,
strides=strides,
data_format=data_format)
inputs = batch_norm_relu(inputs, is_training=is_training)
inputs = dropblock(inputs, is_training=is_training)
inputs = nn_ops.conv2d_fixed_padding(inputs=inputs,
filters=4 * filters,
kernel_size=1,
strides=1,
data_format=data_format)
inputs = batch_norm_relu(inputs, relu=False, is_training=is_training)
inputs = dropblock(inputs, is_training=is_training)
if drop_connect_rate:
inputs = nn_ops.drop_connect(inputs, is_training, drop_connect_rate)
return activation(inputs + shortcut)
def mbconv_block(inputs,
in_filters,
out_filters,
expand_ratio,
strides,
use_projection,
kernel_size=3,
se_ratio=None,
batch_norm_relu=nn_ops.BatchNormRelu(),
dropblock=nn_ops.Dropblock(),
drop_connect_rate=None,
data_format='channels_last',
is_training=False):
"""The bottleneck block with BN and DropBlock after convolutions.
Args:
inputs: a `Tensor` of size `[batch, channels, height, width]`.
in_filters: a `int` number of filters for the input feature map.
out_filters: a `int` number of filters for the output feature map.
expand_ratio: a `int` number as the feature dimension expansion ratio.
strides: a `int` block stride. If greater than 1, this block will ultimately
downsample the input.
use_projection: a `bool` for whether this block should use a projection
shortcut (versus the default identity shortcut). This is usually `True`
for the first block of a block group, which may change the number of
filters and the resolution.
kernel_size: kernel size for the depthwise convolution.
se_ratio: squeeze and excitation ratio.
batch_norm_relu: an operation that is added after convolutions, including a
batch norm layer and an optional relu activation.
dropblock: a drop block layer that is added after convluations. Note that
the default implementation does not apply any drop block.
drop_connect_rate: a 'float' number that specifies the drop connection rate
of the block. Note that the default `None` means no drop connection is
applied.
data_format: a `str` that specifies the data format.
is_training: a `bool` if True, the model is in training mode.
Returns:
The output `Tensor` of the block.
"""
tf.logging.info('-----> Building mbconv block.')
shortcut = inputs
if use_projection:
shortcut = nn_ops.conv2d_fixed_padding(inputs=inputs,
filters=out_filters,
kernel_size=1,
strides=strides,
data_format=data_format)
shortcut = batch_norm_relu(shortcut, is_training=is_training)
shortcut = dropblock(shortcut, is_training=is_training)
# First 1x1 conv for channel expansion.
inputs = nn_ops.conv2d_fixed_padding(inputs=inputs,
filters=in_filters * expand_ratio,
kernel_size=1,
strides=1,
data_format=data_format)
inputs = batch_norm_relu(inputs, is_training=is_training)
inputs = dropblock(inputs, is_training=is_training)
# Second depthwise conv.
inputs = nn_ops.depthwise_conv2d_fixed_padding(inputs=inputs,
kernel_size=kernel_size,
strides=strides,
data_format=data_format)
inputs = batch_norm_relu(inputs, is_training=is_training)
inputs = dropblock(inputs, is_training=is_training)
# Squeeze and excitation.
if se_ratio is not None and se_ratio > 0 and se_ratio <= 1:
inputs = nn_ops.squeeze_excitation(inputs,
in_filters,
se_ratio,
expand_ratio=expand_ratio,
data_format=data_format)
# Third 1x1 conv for reversed bottleneck.
inputs = nn_ops.conv2d_fixed_padding(inputs=inputs,
filters=out_filters,
kernel_size=1,
strides=1,
data_format=data_format)
inputs = batch_norm_relu(inputs, is_training=is_training)
inputs = dropblock(inputs, is_training=is_training)
if drop_connect_rate:
inputs = nn_ops.drop_connect(inputs, is_training, drop_connect_rate)
return tf.add(inputs, shortcut)
def __init__(self,
resnet_depth,
dropblock=nn_ops.Dropblock(),
activation='relu',
batch_norm_activation=nn_ops.BatchNormActivation(),
init_drop_connect_rate=None,
data_format='channels_last',
space_to_depth_block_size=1):
"""ResNet initialization function.
Args:
resnet_depth: `int` depth of ResNet backbone model.
dropblock: a dropblock layer.
activation: activation function. Support 'relu' and 'swish'.
batch_norm_activation: an operation that includes a batch normalization
layer followed by an optional activation layer.
init_drop_connect_rate: a 'float' number that specifies the initial drop
connection rate. Note that the default `None` means no drop connection
is applied.
data_format: `str` either "channels_first" for `[batch, channels, height,
width]` or "channels_last for `[batch, height, width, channels]`.
space_to_depth_block_size: an integer indicates the block size of
space-to-depth convolution for conv0. `0` means use the original conv2d
in ResNet
"""
self._resnet_depth = resnet_depth
self._dropblock = dropblock
if activation == 'relu':
self._activation = tf.nn.relu
elif activation == 'swish':
self._activation = tf.nn.swish
else:
raise ValueError(
'Activation {} not implemented.'.format(activation))
self._batch_norm_activation = batch_norm_activation
self._init_drop_connect_rate = init_drop_connect_rate
self._data_format = data_format
self._space_to_depth_block_size = space_to_depth_block_size
model_params = {
10: {
'block': nn_blocks.residual_block,
'layers': [1, 1, 1, 1]
},
14: {
'block': nn_blocks.bottleneck_block,
'layers': [1, 1, 1, 1]
},
18: {
'block': nn_blocks.residual_block,
'layers': [2, 2, 2, 2]
},
26: {
'block': nn_blocks.bottleneck_block,
'layers': [2, 2, 2, 2]
},
34: {
'block': nn_blocks.residual_block,
'layers': [3, 4, 6, 3]
},
50: {
'block': nn_blocks.bottleneck_block,
'layers': [3, 4, 6, 3]
},
101: {
'block': nn_blocks.bottleneck_block,
'layers': [3, 4, 23, 3]
},
152: {
'block': nn_blocks.bottleneck_block,
'layers': [3, 8, 36, 3]
},
200: {
'block': nn_blocks.bottleneck_block,
'layers': [3, 24, 36, 3]
}
}
if resnet_depth not in model_params:
valid_resnet_depths = ', '.join(
[str(depth) for depth in sorted(model_params.keys())])
raise ValueError(
'The resnet_depth should be in [%s]. Not a valid resnet_depth:'
% (valid_resnet_depths), self._resnet_depth)
params = model_params[resnet_depth]
self._resnet_fn = self.resnet_v1_generator(
params['block'], params['layers'], self._space_to_depth_block_size)
def fused_mbconv_block(inputs,
in_filters,
out_filters,
expand_ratio,
strides,
kernel_size=3,
se_ratio=None,
batch_norm_activation=nn_ops.BatchNormActivation(),
dropblock=nn_ops.Dropblock(),
drop_connect_rate=None,
data_format='channels_last',
is_training=False):
"""The fused bottleneck block with BN and DropBlock after convolutions.
Args:
inputs: a `Tensor` of size `[batch, channels, height, width]`.
in_filters: a `int` number of filters for the input feature map.
out_filters: a `int` number of filters for the output feature map.
expand_ratio: a `int` number as the feature dimension expansion ratio.
strides: a `int` block stride. If greater than 1, this block will ultimately
downsample the input.
kernel_size: kernel size for the depthwise convolution.
se_ratio: squeeze and excitation ratio.
batch_norm_activation: an operation that includes a batch normalization
layer followed by an optional activation layer.
dropblock: a drop block layer that is added after convluations. Note that
the default implementation does not apply any drop block.
drop_connect_rate: a 'float' number that specifies the drop connection rate
of the block. Note that the default `None` means no drop connection is
applied.
data_format: a `str` that specifies the data format.
is_training: a `bool` if True, the model is in training mode.
Returns:
The output `Tensor` of the block.
"""
tf.logging.info('-----> Building fused mbconv block.')
shortcut = inputs
# First 1x1 conv for channel expansion.
inputs = nn_ops.conv2d_fixed_padding(inputs=inputs,
filters=in_filters * expand_ratio,
kernel_size=kernel_size,
strides=strides,
data_format=data_format)
inputs = batch_norm_activation(inputs, is_training=is_training)
inputs = dropblock(inputs, is_training=is_training)
# Squeeze and excitation.
if se_ratio is not None and se_ratio > 0 and se_ratio <= 1:
inputs = nn_ops.squeeze_excitation(inputs,
in_filters,
se_ratio,
expand_ratio=expand_ratio,
data_format=data_format)
# Third 1x1 conv for reversed bottleneck.
inputs = nn_ops.conv2d_fixed_padding(inputs=inputs,
filters=out_filters,
kernel_size=1,
strides=1,
data_format=data_format)
inputs = batch_norm_activation(inputs, relu=False, is_training=is_training)
inputs = dropblock(inputs, is_training=is_training)
if in_filters == out_filters and strides == 1:
if drop_connect_rate:
inputs = nn_ops.drop_connect(inputs, is_training,
drop_connect_rate)
inputs = tf.add(inputs, shortcut)
return inputs
def __init__(self,
min_level=3,
max_level=7,
fpn_feat_dims=256,
num_repeats=7,
use_separable_conv=False,
dropblock=nn_ops.Dropblock(),
block_fn='conv',
block_repeats=1,
activation='swish',
batch_norm_activation=nn_ops.BatchNormActivation(),
init_drop_connect_rate=None):
"""NAS-FPN initialization function.
Args:
min_level: `int` minimum level in NAS-FPN output feature maps.
max_level: `int` maximum level in NAS-FPN output feature maps.
fpn_feat_dims: `int` number of filters in FPN layers.
num_repeats: number of repeats for feature pyramid network.
use_separable_conv: `bool`, if True use separable convolution for
convolution in NAS-FPN layers.
dropblock: a Dropblock layer.
block_fn: `string` representing types of block group support: conv,
bottleneck.
block_repeats: `int` representing the number of repeats per block group
when block group is bottleneck.
activation: activation function. Support 'relu' and 'swish'.
batch_norm_activation: an operation that includes a batch normalization
layer followed by an optional activation layer.
init_drop_connect_rate: a 'float' number that specifies the initial drop
connection rate. Note that the default `None` means no drop connection
is applied.
"""
self._min_level = min_level
self._max_level = max_level
if min_level == 3 and max_level == 7:
model_config = [
3, 1, 1, 3,
3, 0, 1, 5,
4, 0, 0, 6, # Output to level 3.
3, 0, 6, 7, # Output to level 4.
2, 1, 7, 8, # Output to level 5.
0, 1, 6, 9, # Output to level 7.
1, 1, 9, 10] # Output to level 6.
else:
raise ValueError('The NAS-FPN with min level {} and max level {} '
'is not supported.'.format(min_level, max_level))
self._config = Config(model_config, self._min_level, self._max_level)
self._num_repeats = num_repeats
self._fpn_feat_dims = fpn_feat_dims
self._block_fn = block_fn
self._block_repeats = block_repeats
if use_separable_conv:
self._conv2d_op = functools.partial(
tf.layers.separable_conv2d, depth_multiplier=1)
else:
self._conv2d_op = tf.layers.conv2d
self._dropblock = dropblock
if activation == 'relu':
self._activation = tf.nn.relu
elif activation == 'swish':
self._activation = tf.nn.swish
else:
raise ValueError('Activation {} not implemented.'.format(activation))
self._batch_norm_activation = batch_norm_activation
self._init_drop_connect_rate = init_drop_connect_rate
self._resample_feature_map = functools.partial(
resample_feature_map,
target_feat_dims=fpn_feat_dims,
conv2d_op=self._conv2d_op,
batch_norm_activation=batch_norm_activation)
def __init__(self,
min_level=3,
max_level=7,
block_specs=build_block_specs(),
fpn_feat_dims=256,
num_repeats=7,
use_separable_conv=False,
dropblock=nn_ops.Dropblock(),
block_fn='conv',
block_repeats=1,
activation='relu',
batch_norm_activation=nn_ops.BatchNormActivation(
activation='relu'),
init_drop_connect_rate=None,
data_format='channels_last',
use_sum_for_combination=False):
"""NAS-FPN initialization function.
Args:
min_level: `int` minimum level in NAS-FPN output feature maps.
max_level: `int` maximum level in NAS-FPN output feature maps.
block_specs: a list of BlockSpec objects that specifies the SpineNet
network topology. By default, the previously discovered architecture is
used.
fpn_feat_dims: `int` number of filters in FPN layers.
num_repeats: number of repeats for feature pyramid network.
use_separable_conv: `bool`, if True use separable convolution for
convolution in NAS-FPN layers.
dropblock: a Dropblock layer.
block_fn: `string` representing types of block group support: conv,
bottleneck.
block_repeats: `int` representing the number of repeats per block group
when block group is bottleneck.
activation: activation function. Support 'relu' and 'swish'.
batch_norm_activation: an operation that includes a batch normalization
layer followed by an optional activation layer.
init_drop_connect_rate: a 'float' number that specifies the initial drop
connection rate. Note that the default `None` means no drop connection
is applied.
data_format: An optional string from: "channels_last", "channels_first".
Defaults to "channels_last".
use_sum_for_combination: `bool`, if True only 'sum' is used for combining
two nodes.
"""
self._min_level = min_level
self._max_level = max_level
self._block_specs = block_specs
self._fpn_feat_dims = fpn_feat_dims
self._num_repeats = num_repeats
self._block_fn = block_fn
self._block_repeats = block_repeats
if use_separable_conv:
self._conv2d_op = functools.partial(tf.layers.separable_conv2d,
depth_multiplier=1)
else:
self._conv2d_op = tf.layers.conv2d
self._dropblock = dropblock
if activation == 'relu':
self._activation = tf.nn.relu
elif activation == 'swish':
self._activation = tf.nn.swish
else:
raise ValueError(
'Activation {} not implemented.'.format(activation))
self._batch_norm_activation = batch_norm_activation
self._init_drop_connect_rate = init_drop_connect_rate
self._data_format = data_format
self._resample_feature_map = functools.partial(
resample_feature_map,
target_feat_dims=fpn_feat_dims,
conv2d_op=self._conv2d_op,
batch_norm_activation=batch_norm_activation,
data_format=self._data_format)
self._use_sum_for_combination = use_sum_for_combination
