def __init__(self, cfg, input_shape: List[ShapeSpec]):
"""
Arguments:
in_channels (int): number of channels of the input feature
"""
super().__init__()
# TODO: Implement the sigmoid version first.
self.num_classes = cfg.MODEL.FCOS.NUM_CLASSES
self.fpn_strides = cfg.MODEL.FCOS.FPN_STRIDES
head_configs = {
"cls":
(cfg.MODEL.FCOS.NUM_CLS_CONVS, cfg.MODEL.FCOS.USE_DEFORMABLE),
"bbox":
(cfg.MODEL.FCOS.NUM_BOX_CONVS, cfg.MODEL.FCOS.USE_DEFORMABLE),
"share": (cfg.MODEL.FCOS.NUM_SHARE_CONVS, False)
}
norm = None if cfg.MODEL.FCOS.NORM == "none" else cfg.MODEL.FCOS.NORM
self.num_levels = len(input_shape)
in_channels = [s.channels for s in input_shape]
assert len(
set(in_channels)) == 1, "Each level must have the same channel!"
in_channels = in_channels[0]
self.in_channels_to_top_module = in_channels
for head in head_configs:
tower = []
num_convs, use_deformable = head_configs[head]
for i in range(num_convs):
if use_deformable and i == num_convs - 1:
conv_func = DFConv2d
else:
conv_func = nn.Conv2d
tower.append(
conv_func(in_channels,
in_channels,
kernel_size=3,
stride=1,
padding=1,
bias=True))
if norm == "GN":
tower.append(nn.GroupNorm(32, in_channels))
elif norm == "NaiveGN":
tower.append(NaiveGroupNorm(32, in_channels))
elif norm == "BN":
tower.append(
ModuleListDial([
nn.BatchNorm2d(in_channels)
for _ in range(self.num_levels)
]))
elif norm == "SyncBN":
tower.append(
ModuleListDial([
NaiveSyncBatchNorm(in_channels)
for _ in range(self.num_levels)
]))
tower.append(nn.ReLU())
self.add_module('{}_tower'.format(head), nn.Sequential(*tower))
self.cls_logits = nn.Conv2d(in_channels,
self.num_classes,
kernel_size=3,
stride=1,
padding=1)
self.bbox_pred = nn.Conv2d(in_channels,
4,
kernel_size=3,
stride=1,
padding=1)
self.ctrness = nn.Conv2d(in_channels,
1,
kernel_size=3,
stride=1,
padding=1)
if cfg.MODEL.FCOS.USE_SCALE:
self.scales = nn.ModuleList(
[Scale(init_value=1.0) for _ in range(self.num_levels)])
else:
self.scales = None
for modules in [
self.cls_tower, self.bbox_tower, self.share_tower,
self.cls_logits, self.bbox_pred, self.ctrness
]:
for l in modules.modules():
if isinstance(l, nn.Conv2d):
torch.nn.init.normal_(l.weight, std=0.01)
torch.nn.init.constant_(l.bias, 0)
# initialize the bias for focal loss
prior_prob = cfg.MODEL.FCOS.PRIOR_PROB
bias_value = -math.log((1 - prior_prob) / prior_prob)
torch.nn.init.constant_(self.cls_logits.bias, bias_value)
def __init__(self, cfg, input_shape: List[ShapeSpec]):
"""
Arguments:
in_channels (int): number of channels of the input feature
"""
super().__init__()
self.num_classes = cfg.MODEL.DTInst.NUM_CLASSES
self.fpn_strides = cfg.MODEL.DTInst.FPN_STRIDES
self.num_codes = cfg.MODEL.DTInst.NUM_CODE
self.use_gcn_in_mask = cfg.MODEL.DTInst.USE_GCN_IN_MASK
self.gcn_kernel_size = cfg.MODEL.DTInst.GCN_KERNEL_SIZE
self.mask_size = cfg.MODEL.DTInst.MASK_SIZE
self.if_whiten = cfg.MODEL.DTInst.WHITEN
head_configs = {
"cls":
(cfg.MODEL.DTInst.NUM_CLS_CONVS, cfg.MODEL.DTInst.USE_DEFORMABLE),
"bbox":
(cfg.MODEL.DTInst.NUM_BOX_CONVS, cfg.MODEL.DTInst.USE_DEFORMABLE),
"share": (cfg.MODEL.DTInst.NUM_SHARE_CONVS,
cfg.MODEL.DTInst.USE_DEFORMABLE),
"mask":
(cfg.MODEL.DTInst.NUM_MASK_CONVS, cfg.MODEL.DTInst.USE_DEFORMABLE)
}
self.type_deformable = cfg.MODEL.DTInst.TYPE_DEFORMABLE
self.last_deformable = cfg.MODEL.DTInst.LAST_DEFORMABLE
norm = None if cfg.MODEL.DTInst.NORM == "none" else cfg.MODEL.DTInst.NORM
in_channels = [s.channels for s in input_shape]
assert len(
set(in_channels)) == 1, "Each level must have the same channel!"
in_channels = in_channels[0]
for head in head_configs:
tower = []
num_convs, use_deformable = head_configs[head]
for i in range(num_convs):
# conv type.
if use_deformable:
if self.last_deformable:
if i == num_convs - 1:
conv_func = DFConv2d
type_func = self.type_deformable
else:
conv_func = nn.Conv2d
type_func = "Conv2d"
else:
conv_func = DFConv2d
type_func = self.type_deformable
else:
conv_func = nn.Conv2d
type_func = "Conv2d"
# conv operation.
if type_func == "DCNv1":
tower.append(
conv_func(in_channels,
in_channels,
kernel_size=3,
stride=1,
padding=1,
bias=False,
with_modulated_dcn=False))
elif type_func == "DCNv2":
tower.append(
conv_func(in_channels,
in_channels,
kernel_size=3,
stride=1,
padding=1,
bias=False,
with_modulated_dcn=True))
elif type_func == "Conv2d":
tower.append(
conv_func(in_channels,
in_channels,
kernel_size=3,
stride=1,
padding=1,
bias=True))
else:
raise NotImplementedError
# norm.
if norm == "GN":
tower.append(nn.GroupNorm(32, in_channels))
elif norm == "NaiveGN":
tower.append(NaiveGroupNorm(32, in_channels))
# activation.
tower.append(nn.ReLU())
self.add_module('{}_tower'.format(head), nn.Sequential(*tower))
self.cls_logits = nn.Conv2d(in_channels,
self.num_classes,
kernel_size=3,
stride=1,
padding=1)
self.bbox_pred = nn.Conv2d(in_channels,
4,
kernel_size=3,
stride=1,
padding=1)
self.ctrness = nn.Conv2d(in_channels,
1,
kernel_size=3,
stride=1,
padding=1)
# self.residual = nn.Sequential(
# nn.Conv2d(in_channels * 2 + self.mask_size ** 2, in_channels, kernel_size=3, stride=1, padding=1),
# nn.GroupNorm(32, in_channels),
# nn.ReLU(),
# nn.Conv2d(in_channels, in_channels, kernel_size=3, stride=1, padding=1),
# nn.GroupNorm(32, in_channels),
# nn.ReLU(),
# nn.Conv2d(in_channels, self.mask_size ** 2, kernel_size=1, stride=1, padding=0),
# )
self.residual = nn.Sequential(
nn.Conv2d(in_channels * 3,
in_channels,
kernel_size=3,
stride=1,
padding=1),
nn.ReLU(),
nn.Conv2d(in_channels,
in_channels,
kernel_size=3,
stride=1,
padding=1),
nn.ReLU(),
)
# self.residual = nn.Sequential(
# nn.Conv2d(in_channels, in_channels * 2, kernel_size=3, stride=1, padding=1),
# nn.ReLU(),
# nn.Conv2d(in_channels * 2, in_channels, kernel_size=3, stride=1, padding=1),
# nn.ReLU(),
# )
self.code_transform = nn.Conv2d(in_channels,
3,
kernel_size=3,
stride=1,
padding=1)
torch.nn.init.constant_(self.code_transform.bias[0], 1)
torch.nn.init.constant_(self.code_transform.bias[1], 0)
torch.nn.init.constant_(self.code_transform.bias[2], 0.75)
if self.use_gcn_in_mask:
self.mask_pred = GCN(in_channels,
self.num_codes,
k=self.gcn_kernel_size)
else:
self.mask_pred = nn.Conv2d(in_channels,
self.num_codes,
kernel_size=3,
stride=1,
padding=1)
if cfg.MODEL.DTInst.USE_SCALE:
self.scales = nn.ModuleList(
[Scale(init_value=1.0) for _ in self.fpn_strides])
else:
self.scales = None
for modules in [
self.cls_tower, self.bbox_tower, self.share_tower,
self.cls_logits, self.bbox_pred, self.ctrness, self.mask_tower,
self.mask_pred, self.residual
]:
for l in modules.modules():
if isinstance(l, nn.Conv2d):
torch.nn.init.normal_(l.weight, std=0.01)
torch.nn.init.constant_(l.bias, 0)
# initialize the bias for focal loss
prior_prob = cfg.MODEL.DTInst.PRIOR_PROB
bias_value = -math.log((1 - prior_prob) / prior_prob)
torch.nn.init.constant_(self.cls_logits.bias, bias_value)