def __init__(self, w_in, w_out, stride, norm, activation_class, _params):
super().__init__()
self.a = conv2d(w_in, w_out, 3, stride=stride)
self.a_bn = get_norm(norm, w_out)
self.a_af = activation_class()
self.b = conv2d(w_out, w_out, 3)
self.b_bn = get_norm(norm, w_out)
self.b_bn.final_bn = True
def __init__(self, w_in, w_out, stride, norm, activation_class, params):
super().__init__()
w_b = int(round(w_out * params["bot_mul"]))
w_se = int(round(w_in * params["se_r"]))
groups = w_b // params["group_w"]
self.a = conv2d(w_in, w_b, 1)
self.a_bn = get_norm(norm, w_b)
self.a_af = activation_class()
self.b = conv2d(w_b, w_b, 3, stride=stride, groups=groups)
self.b_bn = get_norm(norm, w_b)
self.b_af = activation_class()
self.se = SE(w_b, w_se, activation_class) if w_se else None
self.c = conv2d(w_b, w_out, 1)
self.c_bn = get_norm(norm, w_out)
self.c_bn.final_bn = True
def __init__(self,
input_shape: ShapeSpec,
*,
num_classes,
conv_dims,
conv_norm="",
**kwargs):
"""
NOTE: this interface is experimental.
Args:
input_shape (ShapeSpec): shape of the input feature
num_classes (int): the number of foreground classes (i.e. background is not
included). 1 if using class agnostic prediction.
conv_dims (list[int]): a list of N>0 integers representing the output dimensions
of N-1 conv layers and the last upsample layer.
conv_norm (str or callable): normalization for the conv layers.
See :func:`detectron2.layers.get_norm` for supported types.
"""
super().__init__(**kwargs)
assert len(conv_dims) >= 1, "conv_dims have to be non-empty!"
self.conv_norm_relus = []
cur_channels = input_shape.channels
for k, conv_dim in enumerate(conv_dims[:-1]):
conv = Conv2d(
cur_channels,
conv_dim,
kernel_size=3,
stride=1,
padding=1,
bias=not conv_norm,
norm=get_norm(conv_norm, conv_dim),
activation=nn.ReLU(),
)
self.add_module("mask_fcn{}".format(k + 1), conv)
self.conv_norm_relus.append(conv)
cur_channels = conv_dim
self.deconv = ConvTranspose2d(cur_channels,
conv_dims[-1],
kernel_size=2,
stride=2,
padding=0)
self.add_module("deconv_relu", nn.ReLU())
cur_channels = conv_dims[-1]
self.predictor = Conv2d(cur_channels,
num_classes,
kernel_size=1,
stride=1,
padding=0)
for layer in self.conv_norm_relus + [self.deconv]:
weight_init.c2_msra_fill(layer)
# use normal distribution initialization for mask prediction layer
nn.init.normal_(self.predictor.weight, std=0.001)
if self.predictor.bias is not None:
nn.init.constant_(self.predictor.bias, 0)
def __init__(self, in_channels, out_channels, *, stride=1, norm="BN"):
"""
Args:
in_channels (int): Number of input channels.
out_channels (int): Number of output channels.
stride (int): Stride for the first conv.
norm (str or callable): normalization for all conv layers.
See :func:`layers.get_norm` for supported format.
"""
super().__init__(in_channels, out_channels, stride)
if in_channels != out_channels:
self.shortcut = Conv2d(
in_channels,
out_channels,
kernel_size=1,
stride=stride,
bias=False,
norm=get_norm(norm, out_channels),
)
else:
self.shortcut = None
self.conv1 = Conv2d(
in_channels,
out_channels,
kernel_size=3,
stride=stride,
padding=1,
bias=False,
norm=get_norm(norm, out_channels),
)
self.conv2 = Conv2d(
out_channels,
out_channels,
kernel_size=3,
stride=1,
padding=1,
bias=False,
norm=get_norm(norm, out_channels),
)
for layer in [self.conv1, self.conv2, self.shortcut]:
if layer is not None: # shortcut can be None
weight_init.c2_msra_fill(layer)
def __init__(self, w_in, w_out, stride, norm, activation_class, params):
super().__init__(w_in, w_out, stride)
self.proj, self.bn = None, None
if (w_in != w_out) or (stride != 1):
self.proj = conv2d(w_in, w_out, 1, stride=stride)
self.bn = get_norm(norm, w_out)
self.f = BottleneckTransform(w_in, w_out, stride, norm,
activation_class, params)
self.af = activation_class()
def __init__(
self, input_shape: ShapeSpec, *, conv_dims: List[int], fc_dims: List[int], conv_norm=""
):
"""
NOTE: this interface is experimental.
Args:
input_shape (ShapeSpec): shape of the input feature.
conv_dims (list[int]): the output dimensions of the conv layers
fc_dims (list[int]): the output dimensions of the fc layers
conv_norm (str or callable): normalization for the conv layers.
See :func:`detectron2.layers.get_norm` for supported types.
"""
super().__init__()
assert len(conv_dims) + len(fc_dims) > 0
self._output_size = (input_shape.channels, input_shape.height, input_shape.width)
self.conv_norm_relus = []
for k, conv_dim in enumerate(conv_dims):
conv = Conv2d(
self._output_size[0],
conv_dim,
kernel_size=3,
padding=1,
bias=not conv_norm,
norm=get_norm(conv_norm, conv_dim),
activation=nn.ReLU(),
)
self.add_module("conv{}".format(k + 1), conv)
self.conv_norm_relus.append(conv)
self._output_size = (conv_dim, self._output_size[1], self._output_size[2])
self.fcs = []
for k, fc_dim in enumerate(fc_dims):
if k == 0:
self.add_module("flatten", nn.Flatten())
fc = nn.Linear(int(np.prod(self._output_size)), fc_dim)
self.add_module("fc{}".format(k + 1), fc)
self.add_module("fc_relu{}".format(k + 1), nn.ReLU())
self.fcs.append(fc)
self._output_size = fc_dim
for layer in self.conv_norm_relus:
weight_init.c2_msra_fill(layer)
for layer in self.fcs:
weight_init.c2_xavier_fill(layer)
def __init__(self, in_channels=3, out_channels=64, norm="BN"):
"""
Args:
norm (str or callable): norm after the first conv layer.
See :func:`layers.get_norm` for supported format.
"""
super().__init__(in_channels, out_channels, 4)
self.in_channels = in_channels
self.conv1 = Conv2d(
in_channels,
out_channels,
kernel_size=7,
stride=2,
padding=3,
bias=False,
norm=get_norm(norm, out_channels),
)
weight_init.c2_msra_fill(self.conv1)
def __init__(
self,
in_channels,
out_channels,
*,
bottleneck_channels,
stride=1,
num_groups=1,
norm="BN",
stride_in_1x1=False,
dilation=1,
deform_modulated=False,
deform_num_groups=1,
):
super().__init__(in_channels, out_channels, stride)
self.deform_modulated = deform_modulated
if in_channels != out_channels:
self.shortcut = Conv2d(
in_channels,
out_channels,
kernel_size=1,
stride=stride,
bias=False,
norm=get_norm(norm, out_channels),
)
else:
self.shortcut = None
stride_1x1, stride_3x3 = (stride, 1) if stride_in_1x1 else (1, stride)
self.conv1 = Conv2d(
in_channels,
bottleneck_channels,
kernel_size=1,
stride=stride_1x1,
bias=False,
norm=get_norm(norm, bottleneck_channels),
)
if deform_modulated:
deform_conv_op = ModulatedDeformConv
# offset channels are 2 or 3 (if with modulated) * kernel_size * kernel_size
offset_channels = 27
else:
deform_conv_op = DeformConv
offset_channels = 18
self.conv2_offset = Conv2d(
bottleneck_channels,
offset_channels * deform_num_groups,
kernel_size=3,
stride=stride_3x3,
padding=1 * dilation,
dilation=dilation,
)
self.conv2 = deform_conv_op(
bottleneck_channels,
bottleneck_channels,
kernel_size=3,
stride=stride_3x3,
padding=1 * dilation,
bias=False,
groups=num_groups,
dilation=dilation,
deformable_groups=deform_num_groups,
norm=get_norm(norm, bottleneck_channels),
)
self.conv3 = Conv2d(
bottleneck_channels,
out_channels,
kernel_size=1,
bias=False,
norm=get_norm(norm, out_channels),
)
for layer in [self.conv1, self.conv2, self.conv3, self.shortcut]:
if layer is not None: # shortcut can be None
weight_init.c2_msra_fill(layer)
nn.init.constant_(self.conv2_offset.weight, 0)
nn.init.constant_(self.conv2_offset.bias, 0)
def __init__(
self,
in_channels,
out_channels,
*,
bottleneck_channels,
stride=1,
num_groups=1,
norm="BN",
stride_in_1x1=False,
dilation=1,
):
"""
Args:
bottleneck_channels (int): number of output channels for the 3x3
"bottleneck" conv layers.
num_groups (int): number of groups for the 3x3 conv layer.
norm (str or callable): normalization for all conv layers.
See :func:`layers.get_norm` for supported format.
stride_in_1x1 (bool): when stride>1, whether to put stride in the
first 1x1 convolution or the bottleneck 3x3 convolution.
dilation (int): the dilation rate of the 3x3 conv layer.
"""
super().__init__(in_channels, out_channels, stride)
if in_channels != out_channels:
self.shortcut = Conv2d(
in_channels,
out_channels,
kernel_size=1,
stride=stride,
bias=False,
norm=get_norm(norm, out_channels),
)
else:
self.shortcut = None
# The original MSRA ResNet models have stride in the first 1x1 conv
# The subsequent fb.torch.resnet and Caffe2 ResNe[X]t implementations have
# stride in the 3x3 conv
stride_1x1, stride_3x3 = (stride, 1) if stride_in_1x1 else (1, stride)
self.conv1 = Conv2d(
in_channels,
bottleneck_channels,
kernel_size=1,
stride=stride_1x1,
bias=False,
norm=get_norm(norm, bottleneck_channels),
)
self.conv2 = Conv2d(
bottleneck_channels,
bottleneck_channels,
kernel_size=3,
stride=stride_3x3,
padding=1 * dilation,
bias=False,
groups=num_groups,
dilation=dilation,
norm=get_norm(norm, bottleneck_channels),
)
self.conv3 = Conv2d(
bottleneck_channels,
out_channels,
kernel_size=1,
bias=False,
norm=get_norm(norm, out_channels),
)
for layer in [self.conv1, self.conv2, self.conv3, self.shortcut]:
if layer is not None: # shortcut can be None
weight_init.c2_msra_fill(layer)
def __init__(
self, bottom_up, in_features, out_channels, norm="", top_block=None, fuse_type="sum"
):
"""
Args:
bottom_up (Backbone): module representing the bottom up subnetwork.
Must be a subclass of :class:`Backbone`. The multi-scale feature
maps generated by the bottom up network, and listed in `in_features`,
are used to generate FPN levels.
in_features (list[str]): names of the input feature maps coming
from the backbone to which FPN is attached. For example, if the
backbone produces ["res2", "res3", "res4"], any *contiguous* sublist
of these may be used; order must be from high to low resolution.
out_channels (int): number of channels in the output feature maps.
norm (str): the normalization to use.
top_block (nn.Module or None): if provided, an extra operation will
be performed on the output of the last (smallest resolution)
FPN output, and the result will extend the result list. The top_block
further downsamples the feature map. It must have an attribute
"num_levels", meaning the number of extra FPN levels added by
this block, and "in_feature", which is a string representing
its input feature (e.g., p5).
fuse_type (str): types for fusing the top down features and the lateral
ones. It can be "sum" (default), which sums up element-wise; or "avg",
which takes the element-wise mean of the two.
"""
super(FPN, self).__init__()
assert isinstance(bottom_up, Backbone)
assert in_features, in_features
# Feature map strides and channels from the bottom up network (e.g. ResNet)
input_shapes = bottom_up.output_shape()
strides = [input_shapes[f].stride for f in in_features]
in_channels_per_feature = [input_shapes[f].channels for f in in_features]
_assert_strides_are_log2_contiguous(strides)
lateral_convs = []
output_convs = []
use_bias = norm == ""
for idx, in_channels in enumerate(in_channels_per_feature):
lateral_norm = get_norm(norm, out_channels)
output_norm = get_norm(norm, out_channels)
lateral_conv = Conv2d(
in_channels, out_channels, kernel_size=1, bias=use_bias, norm=lateral_norm
)
output_conv = Conv2d(
out_channels,
out_channels,
kernel_size=3,
stride=1,
padding=1,
bias=use_bias,
norm=output_norm,
)
weight_init.c2_xavier_fill(lateral_conv)
weight_init.c2_xavier_fill(output_conv)
stage = int(math.log2(strides[idx]))
self.add_module("fpn_lateral{}".format(stage), lateral_conv)
self.add_module("fpn_output{}".format(stage), output_conv)
lateral_convs.append(lateral_conv)
output_convs.append(output_conv)
# Place convs into top-down order (from low to high resolution)
# to make the top-down computation in forward clearer.
self.lateral_convs = lateral_convs[::-1]
self.output_convs = output_convs[::-1]
self.top_block = top_block
self.in_features = tuple(in_features)
self.bottom_up = bottom_up
# Return feature names are "p<stage>", like ["p2", "p3", ..., "p6"]
self._out_feature_strides = {"p{}".format(int(math.log2(s))): s for s in strides}
# top block output feature maps.
if self.top_block is not None:
for s in range(stage, stage + self.top_block.num_levels):
self._out_feature_strides["p{}".format(s + 1)] = 2 ** (s + 1)
self._out_features = list(self._out_feature_strides.keys())
self._out_feature_channels = {k: out_channels for k in self._out_features}
self._size_divisibility = strides[-1]
assert fuse_type in {"avg", "sum"}
self._fuse_type = fuse_type
def __init__(self, w_in, w_out, norm, activation_class):
super().__init__(w_in, w_out, 2)
self.conv = conv2d(w_in, w_out, 3, stride=2)
self.bn = get_norm(norm, w_out)
self.af = activation_class()
def __init__(
self,
input_shape: Dict[str, ShapeSpec],
*,
num_classes: int,
conv_dims: int,
common_stride: int,
loss_weight: float = 1.0,
norm: Optional[Union[str, Callable]] = None,
ignore_value: int = -1,
):
"""
NOTE: this interface is experimental.
Args:
input_shape: shapes (channels and stride) of the input features
num_classes: number of classes to predict
conv_dims: number of output channels for the intermediate conv layers.
common_stride: the common stride that all features will be upscaled to
loss_weight: loss weight
norm (str or callable): normalization for all conv layers
ignore_value: category id to be ignored during training.
"""
super().__init__()
input_shape = sorted(input_shape.items(), key=lambda x: x[1].stride)
self.in_features = [k for k, v in input_shape]
feature_strides = [v.stride for k, v in input_shape]
feature_channels = [v.channels for k, v in input_shape]
self.ignore_value = ignore_value
self.common_stride = common_stride
self.loss_weight = loss_weight
self.scale_heads = []
for in_feature, stride, channels in zip(self.in_features,
feature_strides,
feature_channels):
head_ops = []
head_length = max(
1, int(np.log2(stride) - np.log2(self.common_stride)))
for k in range(head_length):
norm_module = get_norm(norm, conv_dims)
conv = Conv2d(
channels if k == 0 else conv_dims,
conv_dims,
kernel_size=3,
stride=1,
padding=1,
bias=not norm,
norm=norm_module,
activation=F.relu,
)
weight_init.c2_msra_fill(conv)
head_ops.append(conv)
if stride != self.common_stride:
head_ops.append(
nn.Upsample(scale_factor=2,
mode="bilinear",
align_corners=False))
self.scale_heads.append(nn.Sequential(*head_ops))
self.add_module(in_feature, self.scale_heads[-1])
self.predictor = Conv2d(conv_dims,
num_classes,
kernel_size=1,
stride=1,
padding=0)
weight_init.c2_msra_fill(self.predictor)