def __init__(self,
in_channels,
out_channels,
strides,
cardinality,
bottleneck_width,
data_format="channels_last",
**kwargs):
super(SEResNeXtUnit, self).__init__(**kwargs)
self.resize_identity = (in_channels != out_channels) or (strides != 1)
self.body = ResNeXtBottleneck(
in_channels=in_channels,
out_channels=out_channels,
strides=strides,
cardinality=cardinality,
bottleneck_width=bottleneck_width,
data_format=data_format,
name="body")
self.se = SEBlock(
channels=out_channels,
data_format=data_format,
name="se")
if self.resize_identity:
self.identity_conv = conv1x1_block(
in_channels=in_channels,
out_channels=out_channels,
strides=strides,
activation=None,
data_format=data_format,
name="identity_conv")
self.activ = nn.ReLU()
def __init__(self,
in_channels,
out_channels,
strides,
bottleneck,
conv1_stride,
data_format="channels_last",
**kwargs):
super(SEResUnit, self).__init__(**kwargs)
self.resize_identity = (in_channels != out_channels) or (strides != 1)
if bottleneck:
self.body = ResBottleneck(in_channels=in_channels,
out_channels=out_channels,
strides=strides,
conv1_stride=conv1_stride,
data_format=data_format,
name="body")
else:
self.body = ResBlock(in_channels=in_channels,
out_channels=out_channels,
strides=strides,
data_format=data_format,
name="body")
self.se = SEBlock(channels=out_channels)
if self.resize_identity:
self.identity_conv = conv1x1_block(in_channels=in_channels,
out_channels=out_channels,
strides=strides,
activation=None,
data_format=data_format,
name="identity_conv")
self.activ = nn.ReLU()
def __init__(self,
in_channels,
out_channels,
strides,
bn_eps,
activation,
tf_mode,
data_format="channels_last",
**kwargs):
super(EffiDwsConvUnit, self).__init__(**kwargs)
self.tf_mode = tf_mode
self.data_format = data_format
self.residual = (in_channels == out_channels) and (strides == 1)
self.dw_conv = dwconv3x3_block(in_channels=in_channels,
out_channels=in_channels,
padding=(0 if tf_mode else 1),
bn_eps=bn_eps,
activation=activation,
data_format=data_format,
name="dw_conv")
self.se = SEBlock(channels=in_channels,
reduction=4,
mid_activation=activation,
data_format=data_format,
name="se")
self.pw_conv = conv1x1_block(in_channels=in_channels,
out_channels=out_channels,
bn_eps=bn_eps,
activation=None,
data_format=data_format,
name="pw_conv")
def __init__(self,
out_channels,
use_bn,
activation,
data_format="channels_last",
**kwargs):
super(IbpPreBlock, self).__init__(**kwargs)
self.conv1 = conv3x3_block(in_channels=out_channels,
out_channels=out_channels,
use_bias=(not use_bn),
use_bn=use_bn,
activation=activation,
data_format=data_format,
name="conv1")
self.conv2 = conv3x3_block(in_channels=out_channels,
out_channels=out_channels,
use_bias=(not use_bn),
use_bn=use_bn,
activation=activation,
data_format=data_format,
name="conv2")
self.se = SEBlock(channels=out_channels,
use_conv=False,
mid_activation=activation,
data_format=data_format,
name="se")
def __init__(self,
in_channels,
out_channels,
kernel_size,
strides,
exp_factor,
se_factor,
bn_eps,
activation,
tf_mode,
data_format="channels_last",
**kwargs):
super(EffiInvResUnit, self).__init__(**kwargs)
self.kernel_size = kernel_size
self.strides = strides
self.tf_mode = tf_mode
self.data_format = data_format
self.residual = (in_channels == out_channels) and (strides == 1)
self.use_se = se_factor > 0
mid_channels = in_channels * exp_factor
dwconv_block_fn = dwconv3x3_block if kernel_size == 3 else (
dwconv5x5_block if kernel_size == 5 else None)
self.conv1 = conv1x1_block(in_channels=in_channels,
out_channels=mid_channels,
bn_eps=bn_eps,
activation=activation,
data_format=data_format,
name="conv1")
self.conv2 = dwconv_block_fn(in_channels=mid_channels,
out_channels=mid_channels,
strides=strides,
padding=(0 if tf_mode else
(kernel_size // 2)),
bn_eps=bn_eps,
activation=activation,
data_format=data_format,
name="conv2")
if self.use_se:
self.se = SEBlock(channels=mid_channels,
reduction=(exp_factor * se_factor),
mid_activation=activation,
data_format=data_format,
name="se")
self.conv3 = conv1x1_block(in_channels=mid_channels,
out_channels=out_channels,
bn_eps=bn_eps,
activation=None,
data_format=data_format,
name="conv3")
def __init__(self,
in_channels,
out_channels,
exp_channels,
strides,
use_kernel3,
activation,
use_se,
data_format="channels_last",
**kwargs):
super(MobileNetV3Unit, self).__init__(**kwargs)
assert (exp_channels >= out_channels)
self.residual = (in_channels == out_channels) and (strides == 1)
self.use_se = use_se
self.use_exp_conv = exp_channels != out_channels
mid_channels = exp_channels
if self.use_exp_conv:
self.exp_conv = conv1x1_block(in_channels=in_channels,
out_channels=mid_channels,
activation=activation,
data_format=data_format,
name="exp_conv")
if use_kernel3:
self.conv1 = dwconv3x3_block(in_channels=mid_channels,
out_channels=mid_channels,
strides=strides,
activation=activation,
data_format=data_format,
name="conv1")
else:
self.conv1 = dwconv5x5_block(in_channels=mid_channels,
out_channels=mid_channels,
strides=strides,
activation=activation,
data_format=data_format,
name="conv1")
if self.use_se:
self.se = SEBlock(channels=mid_channels,
reduction=4,
round_mid=True,
out_activation="hsigmoid",
data_format=data_format,
name="se")
self.conv2 = conv1x1_block(in_channels=mid_channels,
out_channels=out_channels,
activation=None,
data_format=data_format,
name="conv2")
def __init__(self,
in_channels,
out_channels,
strides=1,
use_kernel3=True,
exp_factor=1,
se_factor=0,
use_skip=True,
activation="relu",
data_format="channels_last",
**kwargs):
super(DwsExpSEResUnit, self).__init__(**kwargs)
assert (exp_factor >= 1)
self.residual = (in_channels == out_channels) and (strides
== 1) and use_skip
self.use_exp_conv = exp_factor > 1
self.use_se = se_factor > 0
mid_channels = exp_factor * in_channels
dwconv_block_fn = dwconv3x3_block if use_kernel3 else dwconv5x5_block
if self.use_exp_conv:
self.exp_conv = conv1x1_block(in_channels=in_channels,
out_channels=mid_channels,
activation=activation,
data_format=data_format,
name="exp_conv")
self.dw_conv = dwconv_block_fn(in_channels=mid_channels,
out_channels=mid_channels,
strides=strides,
activation=activation,
data_format=data_format,
name="dw_conv")
if self.use_se:
self.se = SEBlock(channels=mid_channels,
reduction=(exp_factor * se_factor),
round_mid=False,
mid_activation=activation,
data_format=data_format,
name="se")
self.pw_conv = conv1x1_block(in_channels=mid_channels,
out_channels=out_channels,
activation=None,
data_format=data_format,
name="pw_conv")
def __init__(self,
in_channels,
out_channels,
use_se,
data_format="channels_last",
**kwargs):
super(MobileNetV3FinalBlock, self).__init__(**kwargs)
self.use_se = use_se
self.conv = conv1x1_block(in_channels=in_channels,
out_channels=out_channels,
activation="hswish",
data_format=data_format,
name="conv")
if self.use_se:
self.se = SEBlock(channels=out_channels,
reduction=4,
round_mid=True,
out_activation="hsigmoid",
data_format=data_format,
name="se")
def __init__(self,
in_channels,
out_channels,
strides,
exp_factor,
se_factor,
mid_from_in,
use_skip,
bn_eps,
activation,
data_format="channels_last",
**kwargs):
super(EffiEdgeResUnit, self).__init__(**kwargs)
self.residual = (in_channels == out_channels) and (strides
== 1) and use_skip
self.use_se = se_factor > 0
mid_channels = in_channels * exp_factor if mid_from_in else out_channels * exp_factor
self.conv1 = conv3x3_block(in_channels=in_channels,
out_channels=mid_channels,
bn_eps=bn_eps,
activation=activation,
data_format=data_format,
name="conv1")
if self.use_se:
self.se = SEBlock(channels=mid_channels,
reduction=(exp_factor * se_factor),
mid_activation=activation,
data_format=data_format,
name="se")
self.conv2 = conv1x1_block(in_channels=mid_channels,
out_channels=out_channels,
strides=strides,
bn_eps=bn_eps,
activation=None,
data_format=data_format,
name="conv2")
def __init__(self,
in_channels,
out_channels,
strides,
use_kernel3,
exp_factor,
use_se,
bn_use_global_stats=False,
**kwargs):
super(GhostExpBlock, self).__init__(**kwargs)
self.use_dw_conv = (strides != 1)
self.use_se = use_se
mid_channels = int(math.ceil(exp_factor * in_channels))
with self.name_scope():
self.exp_conv = GhostConvBlock(
in_channels=in_channels,
out_channels=mid_channels,
bn_use_global_stats=bn_use_global_stats)
if self.use_dw_conv:
dw_conv_class = dwconv3x3_block if use_kernel3 else dwconv5x5_block
self.dw_conv = dw_conv_class(
in_channels=mid_channels,
out_channels=mid_channels,
strides=strides,
bn_use_global_stats=bn_use_global_stats,
activation=None)
if self.use_se:
self.se = SEBlock(channels=mid_channels,
reduction=4,
out_activation=HSigmoid())
self.pw_conv = GhostConvBlock(
in_channels=mid_channels,
out_channels=out_channels,
bn_use_global_stats=bn_use_global_stats,
activation=None)
def __init__(self,
in_channels,
out_channels,
strides,
use_kernel3,
exp_factor,
use_se,
data_format="channels_last",
**kwargs):
super(GhostExpBlock, self).__init__(**kwargs)
self.use_dw_conv = (strides != 1)
self.use_se = use_se
mid_channels = int(math.ceil(exp_factor * in_channels))
self.exp_conv = GhostConvBlock(in_channels=in_channels,
out_channels=mid_channels,
name="exp_conv")
if self.use_dw_conv:
dw_conv_class = dwconv3x3_block if use_kernel3 else dwconv5x5_block
self.dw_conv = dw_conv_class(in_channels=mid_channels,
out_channels=mid_channels,
strides=strides,
activation=None,
data_format=data_format,
name="dw_conv")
if self.use_se:
self.se = SEBlock(channels=mid_channels,
reduction=4,
out_activation=GhostHSigmoid(),
data_format=data_format,
name="se")
self.pw_conv = GhostConvBlock(in_channels=mid_channels,
out_channels=out_channels,
activation=None,
data_format=data_format,
name="pw_conv")
def __init__(self,
in_channels,
out_channels,
downsample,
use_se,
use_residual,
data_format="channels_last",
**kwargs):
super(ShuffleUnit, self).__init__(**kwargs)
self.data_format = data_format
self.downsample = downsample
self.use_se = use_se
self.use_residual = use_residual
mid_channels = out_channels // 2
self.compress_conv1 = conv1x1(
in_channels=(in_channels if self.downsample else mid_channels),
out_channels=mid_channels,
data_format=data_format,
name="compress_conv1")
self.compress_bn1 = BatchNorm(
# in_channels=mid_channels,
data_format=data_format,
name="compress_bn1")
self.dw_conv2 = depthwise_conv3x3(
channels=mid_channels,
strides=(2 if self.downsample else 1),
data_format=data_format,
name="dw_conv2")
self.dw_bn2 = BatchNorm(
# in_channels=mid_channels,
data_format=data_format,
name="dw_bn2")
self.expand_conv3 = conv1x1(in_channels=mid_channels,
out_channels=mid_channels,
data_format=data_format,
name="expand_conv3")
self.expand_bn3 = BatchNorm(
# in_channels=mid_channels,
data_format=data_format,
name="expand_bn3")
if self.use_se:
self.se = SEBlock(channels=mid_channels,
data_format=data_format,
name="se")
if downsample:
self.dw_conv4 = depthwise_conv3x3(channels=in_channels,
strides=2,
data_format=data_format,
name="dw_conv4")
self.dw_bn4 = BatchNorm(
# in_channels=in_channels,
data_format=data_format,
name="dw_bn4")
self.expand_conv5 = conv1x1(in_channels=in_channels,
out_channels=mid_channels,
data_format=data_format,
name="expand_conv5")
self.expand_bn5 = BatchNorm(
# in_channels=mid_channels,
data_format=data_format,
name="expand_bn5")
self.activ = nn.ReLU()
self.c_shuffle = ChannelShuffle(channels=out_channels,
groups=2,
data_format=data_format,
name="c_shuffle")
def __init__(self,
in_channels,
out_channels,
downsample,
use_se,
use_residual,
shuffle_group_first,
data_format="channels_last",
**kwargs):
super(ShuffleUnit, self).__init__(**kwargs)
self.data_format = data_format
self.downsample = downsample
self.use_se = use_se
self.use_residual = use_residual
mid_channels = out_channels // 2
in_channels2 = in_channels // 2
assert (in_channels % 2 == 0)
y2_in_channels = (in_channels if downsample else in_channels2)
y2_out_channels = out_channels - y2_in_channels
self.conv1 = conv1x1_block(in_channels=y2_in_channels,
out_channels=mid_channels,
data_format=data_format,
name="conv1")
self.dconv = dwconv3x3_block(in_channels=mid_channels,
out_channels=mid_channels,
strides=(2 if self.downsample else 1),
activation=None,
data_format=data_format,
name="dconv")
self.conv2 = conv1x1_block(in_channels=mid_channels,
out_channels=y2_out_channels,
data_format=data_format,
name="conv2")
if self.use_se:
self.se = SEBlock(channels=y2_out_channels,
data_format=data_format,
name="se")
if downsample:
self.shortcut_dconv = dwconv3x3_block(in_channels=in_channels,
out_channels=in_channels,
strides=2,
activation=None,
data_format=data_format,
name="shortcut_dconv")
self.shortcut_conv = conv1x1_block(in_channels=in_channels,
out_channels=in_channels,
data_format=data_format,
name="shortcut_conv")
if shuffle_group_first:
self.c_shuffle = ChannelShuffle(channels=out_channels,
groups=2,
data_format=data_format,
name="c_shuffle")
else:
self.c_shuffle = ChannelShuffle2(channels=out_channels,
groups=2,
data_format=data_format,
name="c_shuffle")
def __init__(self,
in_channels,
out_channels,
strides,
exp_kernel_count,
conv1_kernel_count,
conv2_kernel_count,
exp_factor,
se_factor,
activation,
data_format="channels_last",
**kwargs):
super(MixUnit, self).__init__(**kwargs)
assert (exp_factor >= 1)
assert (se_factor >= 0)
self.residual = (in_channels == out_channels) and (strides == 1)
self.use_se = se_factor > 0
mid_channels = exp_factor * in_channels
self.use_exp_conv = exp_factor > 1
if self.use_exp_conv:
if exp_kernel_count == 1:
self.exp_conv = conv1x1_block(in_channels=in_channels,
out_channels=mid_channels,
activation=activation,
data_format=data_format,
name="exp_conv")
else:
self.exp_conv = mixconv1x1_block(in_channels=in_channels,
out_channels=mid_channels,
kernel_count=exp_kernel_count,
activation=activation,
data_format=data_format,
name="exp_conv")
if conv1_kernel_count == 1:
self.conv1 = dwconv3x3_block(in_channels=mid_channels,
out_channels=mid_channels,
strides=strides,
activation=activation,
data_format=data_format,
name="conv1")
else:
self.conv1 = MixConvBlock(
in_channels=mid_channels,
out_channels=mid_channels,
kernel_size=[3 + 2 * i for i in range(conv1_kernel_count)],
strides=strides,
padding=[1 + i for i in range(conv1_kernel_count)],
groups=mid_channels,
activation=activation,
data_format=data_format,
name="conv1")
if self.use_se:
self.se = SEBlock(channels=mid_channels,
reduction=(exp_factor * se_factor),
round_mid=False,
mid_activation=activation,
data_format=data_format,
name="se")
if conv2_kernel_count == 1:
self.conv2 = conv1x1_block(in_channels=mid_channels,
out_channels=out_channels,
activation=None,
data_format=data_format,
name="conv2")
else:
self.conv2 = mixconv1x1_block(in_channels=mid_channels,
out_channels=out_channels,
kernel_count=conv2_kernel_count,
activation=None,
data_format=data_format,
name="conv2")