def __init__(self,
in_chs,
out_chs,
dilation=1,
bottle_ratio=0.5,
groups=1,
act_layer=nn.ReLU,
norm_layer=nn.BatchNorm2d,
attn_layer=None,
aa_layer=None,
drop_block=None,
drop_path=None):
super(DarkBlock_ACON, self).__init__()
mid_chs = int(round(out_chs * bottle_ratio))
ckwargs = dict(act_layer=act_layer,
norm_layer=norm_layer,
aa_layer=aa_layer,
drop_block=drop_block)
self.conv1 = ConvBnAct(in_chs, mid_chs, kernel_size=1, **ckwargs)
self.conv2 = Conv3x3_ACON(mid_chs,
out_chs,
kernel_size=3,
dilation=dilation,
groups=groups)
self.attn = create_attn(attn_layer, channels=out_chs)
self.drop_path = drop_path
def __init__(self, inplanes, planes, stride=1, downsample=None, cardinality=1, base_width=64,
reduce_first=1, dilation=1, first_dilation=None, act_layer=nn.ReLU, norm_layer=nn.BatchNorm2d,
attn_layer=None, drop_block=None, drop_path=None):
super(Bottleneck, self).__init__()
width = int(math.floor(planes * (base_width / 64)) * cardinality)
first_planes = width // reduce_first
outplanes = planes * self.expansion
first_dilation = first_dilation or dilation
self.conv1 = nn.Conv2d(inplanes, first_planes, kernel_size=1, bias=False)
self.bn1 = norm_layer(first_planes)
self.act1 = act_layer(inplace=True)
self.conv2 = nn.Conv2d(
first_planes, width, kernel_size=3, stride=stride,
padding=first_dilation, dilation=first_dilation, groups=cardinality, bias=False)
self.bn2 = norm_layer(width)
self.act2 = act_layer(inplace=True)
self.conv3 = nn.Conv2d(width, outplanes, kernel_size=1, bias=False)
self.bn3 = norm_layer(outplanes)
self.se = create_attn(attn_layer, outplanes)
self.act3 = act_layer(inplace=True)
self.downsample = downsample
self.stride = stride
self.dilation = dilation
self.drop_block = drop_block
self.drop_path = drop_path
def __init__(self, inplanes, planes, stride=1, downsample=None, cardinality=1, base_width=64,
reduce_first=1, dilation=1, first_dilation=None, act_layer=nn.ReLU, norm_layer=nn.BatchNorm2d,
attn_layer=None, drop_block=None, drop_path=None):
super(BasicBlock, self).__init__()
assert cardinality == 1, 'BasicBlock only supports cardinality of 1'
assert base_width == 64, 'BasicBlock doest not support changing base width'
first_planes = planes // reduce_first
outplanes = planes * self.expansion
first_dilation = first_dilation or dilation
self.conv1 = nn.Conv2d(
inplanes, first_planes, kernel_size=3, stride=stride, padding=first_dilation,
dilation=first_dilation, bias=False)
self.bn1 = norm_layer(first_planes)
self.act1 = act_layer(inplace=True)
self.conv2 = nn.Conv2d(
first_planes, outplanes, kernel_size=3, padding=dilation, dilation=dilation, bias=False)
self.bn2 = norm_layer(outplanes)
self.se = create_attn(attn_layer, outplanes)
self.act2 = act_layer(inplace=True)
self.downsample = downsample
self.stride = stride
self.dilation = dilation
self.drop_block = drop_block
self.drop_path = drop_path
def __init__(self, inplanes, planes, stride=1, cardinality=1, base_width=64, reduce_first=1, dilation=1,
first_dilation=None, act_layer=nn.ReLU, norm_layer=nn.BatchNorm2d, attn_layer=None,
aa_layer=None, drop_block=None, drop_path=None, fmap_size=56, up_inplanes=None):
super(ResConv, self).__init__()
width = int(math.floor(planes * (base_width / 64)) * cardinality)
first_planes = width // reduce_first
outplanes = planes * self.expansion
first_dilation = first_dilation or dilation
use_aa = aa_layer is not None and (stride == 2 or first_dilation != dilation)
self.fmap_size = fmap_size
if up_inplanes is not None:
self.downsample_d = nn.Sequential(
nn.Conv2d(up_inplanes, outplanes, 3, stride=2, padding=1, bias=False),
nn.BatchNorm2d(outplanes),
act_layer(inplace=True))
if inplanes != outplanes:
self.downsample = nn.Sequential(
nn.Conv2d(inplanes, outplanes, 1, stride=1, padding=0, bias=False),
nn.BatchNorm2d(outplanes),
act_layer(inplace=True))
else:
self.downsample = nn.Identity()
if up_inplanes is not None:
self.conv1_d = nn.Conv2d(up_inplanes, first_planes, kernel_size=1, bias=False)
self.peg_d = PEG(first_planes, stride=2)
self.bn1_d = norm_layer(first_planes)
self.conv1 = nn.Conv2d(inplanes, first_planes, kernel_size=1, bias=False)
self.peg = PEG(first_planes, stride=1)
self.bn1 = norm_layer(first_planes)
self.act1 = act_layer(inplace=True)
self.conv2 = nn.Conv2d(
first_planes, width, kernel_size=3, stride=1 if use_aa else stride,
padding=first_dilation, dilation=first_dilation, groups=cardinality, bias=False)
self.bn2 = norm_layer(width)
self.act2 = act_layer(inplace=True)
self.aa = aa_layer(channels=width, stride=stride) if use_aa else None
self.conv3 = nn.Conv2d(width, outplanes, kernel_size=1, bias=False)
self.bn3 = norm_layer(outplanes)
self.se = create_attn(attn_layer, outplanes)
self.act3 = act_layer(inplace=True)
self.stride = stride
self.dilation = dilation
self.drop_block = drop_block
self.drop_path = drop_path
self.inc = inplanes
self.zero_init_last_bn()
def __init__(self,
inplanes,
planes,
stride=1,
cardinality=1,
base_width=64,
reduce_first=1,
dilation=1,
first_dilation=None,
act_layer=nn.ReLU,
norm_layer=nn.BatchNorm2d,
attn_layer=None,
aa_layer=None,
drop_block=None,
drop_path=None,
avg_down=False):
super(ResConv, self).__init__()
width = int(math.floor(planes * (base_width / 64)) * cardinality)
first_planes = width // reduce_first
outplanes = planes * self.expansion
first_dilation = first_dilation or dilation
use_aa = aa_layer is not None and (stride == 2
or first_dilation != dilation)
# act_layer = nn.SiLU
if avg_down and (stride == 2 or inplanes != outplanes):
avg_stride = stride if dilation == 1 else 1
if stride == 1 and dilation == 1:
pool = nn.Identity()
else:
avg_pool_fn = AvgPool2dSame if avg_stride == 1 and dilation > 1 else nn.AvgPool2d
pool = avg_pool_fn(2,
avg_stride,
ceil_mode=True,
count_include_pad=False)
self.downsample = nn.Sequential(*[
pool,
nn.Conv2d(
inplanes, outplanes, 1, stride=1, padding=0, bias=False),
norm_layer(outplanes)
])
else:
if stride == 2:
self.downsample = nn.Sequential(
nn.Conv2d(inplanes,
outplanes,
3,
stride=2,
padding=1,
bias=False), norm_layer(outplanes),
act_layer(inplace=True))
elif inplanes != outplanes:
self.downsample = nn.Sequential(
nn.Conv2d(inplanes,
outplanes,
1,
stride=1,
padding=0,
bias=False), norm_layer(outplanes),
act_layer(inplace=True))
else:
self.downsample = nn.Identity()
self.conv1 = nn.Conv2d(inplanes,
first_planes,
kernel_size=1,
bias=False)
self.peg = PEG(first_planes, stride=stride)
self.bn1 = norm_layer(first_planes)
self.act1 = act_layer(inplace=True)
self.conv2 = nn.Conv2d(first_planes,
width,
kernel_size=3,
stride=1,
padding=first_dilation,
dilation=first_dilation,
groups=cardinality,
bias=False)
self.bn2 = norm_layer(width)
self.act2 = act_layer(inplace=True)
self.aa = aa_layer(channels=width, stride=stride) if use_aa else None
self.conv3 = nn.Conv2d(width, outplanes, kernel_size=1, bias=False)
self.bn3 = norm_layer(outplanes)
if attn_layer == 'se':
self.se = create_attn(attn_layer, outplanes)
else:
self.se = None
self.act3 = act_layer(inplace=True)
# self.downsample = downsample
self.stride = stride
self.dilation = dilation
self.drop_block = drop_block
self.drop_path = drop_path
self.inc = inplanes