def build_trident_resnet_backbone(cfg, input_shape):
"""
Create a ResNet instance from config for TridentNet.
Returns:
ResNet: a :class:`ResNet` instance.
"""
# need registration of new blocks/stems?
norm = cfg.MODEL.RESNETS.NORM
activation = cfg.MODEL.RESNETS.ACTIVATION
deep_stem = cfg.MODEL.RESNETS.DEEP_STEM
if not deep_stem:
assert getattr(cfg.MODEL.RESNETS, "RADIX", 1) <= 1, \
"cfg.MODEL.RESNETS.RADIX > 1: {}".format(cfg.MODEL.RESNETS.RADIX)
stem = BasicStem(
in_channels=input_shape.channels,
out_channels=cfg.MODEL.RESNETS.STEM_OUT_CHANNELS,
norm=norm,
activation=activation,
deep_stem=deep_stem,
)
freeze_at = cfg.MODEL.BACKBONE.FREEZE_AT
if freeze_at >= 1:
for p in stem.parameters():
p.requires_grad = False
stem = FrozenBatchNorm2d.convert_frozen_batchnorm(stem)
# fmt: off
out_features = cfg.MODEL.RESNETS.OUT_FEATURES
depth = cfg.MODEL.RESNETS.DEPTH
num_groups = cfg.MODEL.RESNETS.NUM_GROUPS
width_per_group = cfg.MODEL.RESNETS.WIDTH_PER_GROUP
bottleneck_channels = num_groups * width_per_group
in_channels = cfg.MODEL.RESNETS.STEM_OUT_CHANNELS
out_channels = cfg.MODEL.RESNETS.RES2_OUT_CHANNELS
stride_in_1x1 = cfg.MODEL.RESNETS.STRIDE_IN_1X1
res5_dilation = cfg.MODEL.RESNETS.RES5_DILATION
num_branch = cfg.MODEL.TRIDENT.NUM_BRANCH
branch_dilations = cfg.MODEL.TRIDENT.BRANCH_DILATIONS
trident_stage = cfg.MODEL.TRIDENT.TRIDENT_STAGE
test_branch_idx = cfg.MODEL.TRIDENT.TEST_BRANCH_IDX
# fmt: on
assert res5_dilation in {
1, 2
}, "res5_dilation cannot be {}.".format(res5_dilation)
num_blocks_per_stage = {
18: [2, 2, 2, 2],
34: [3, 4, 6, 3],
50: [3, 4, 6, 3],
101: [3, 4, 23, 3],
152: [3, 8, 36, 3],
}[depth]
res_stage_idx = {"res2": 2, "res3": 3, "res4": 4, "res5": 5, "linear": 5}
out_stage_idx = [res_stage_idx[f] for f in out_features]
trident_stage_idx = res_stage_idx[trident_stage]
max_stage_idx = max(out_stage_idx)
deform_on_per_stage = getattr(cfg.MODEL.RESNETS, "DEFORM_ON_PER_STAGE",
[False] * (max_stage_idx - 1))
stages = []
if depth in [18, 34]:
assert out_channels == 64, "Must set MODEL.RESNETS.RES2_OUT_CHANNELS = 64 for R18/R34"
assert not any(
deform_on_per_stage
), "MODEL.RESNETS.DEFORM_ON_PER_STAGE unsupported for R18/R34"
assert res5_dilation == 1, "Must set MODEL.RESNETS.RES5_DILATION = 1 for R18/R34"
assert num_groups == 1, "Must set MODEL.RESNETS.NUM_GROUPS = 1 for R18/R34"
for idx, stage_idx in enumerate(range(2, max_stage_idx + 1)):
dilation = res5_dilation if stage_idx == 5 else 1
first_stride = 1 if idx == 0 or (stage_idx == 5
and dilation == 2) else 2
stage_kargs = {
"num_blocks": num_blocks_per_stage[idx],
"first_stride": first_stride,
"in_channels": in_channels,
"out_channels": out_channels,
"norm": norm,
"activation": activation,
}
# Use BasicBlock for R18 and R34.
if depth in [18, 34]:
stage_kargs["block_class"] = BasicBlock
else:
stage_kargs["bottleneck_channels"] = bottleneck_channels
stage_kargs["stride_in_1x1"] = stride_in_1x1
stage_kargs["dilation"] = dilation
stage_kargs["num_groups"] = num_groups
if stage_idx == trident_stage_idx:
assert not deform_on_per_stage[
idx], "Not support deformable conv in Trident blocks yet."
stage_kargs["block_class"] = TridentBottleneckBlock
stage_kargs["num_branch"] = num_branch
stage_kargs["dilations"] = branch_dilations
stage_kargs["test_branch_idx"] = test_branch_idx
stage_kargs.pop("dilation")
elif deform_on_per_stage[idx]:
stage_kargs["block_class"] = DeformBottleneckBlock
# Use True to use modulated deform_conv (DeformableV2);
# Use False for DeformableV1.
stage_kargs[
"deform_modulated"] = cfg.MODEL.RESNETS.DEFORM_MODULATED
# Number of groups in deformable conv.
stage_kargs[
"deform_num_groups"] = cfg.MODEL.RESNETS.DEFORM_NUM_GROUPS
elif hasattr(cfg.MODEL.RESNETS, "RADIX"):
# Radix in ResNeSt
radix = cfg.MODEL.RESNETS.RADIX
# Apply avg after conv2 in the BottleBlock
# When AVD=True, the STRIDE_IN_1X1 should be False
avd = cfg.MODEL.RESNETS.AVD or (radix > 1)
# Apply avg_down to the downsampling layer for residual path
avg_down = cfg.MODEL.RESNETS.AVG_DOWN or (radix > 1)
# Bottleneck_width in ResNeSt
bottleneck_width = cfg.MODEL.RESNETS.BOTTLENECK_WIDTH
stage_kargs["block_class"] = AVDBottleneckBlock
stage_kargs["avd"] = avd
stage_kargs["avg_down"] = avg_down
stage_kargs["radix"] = radix
stage_kargs["bottleneck_width"] = bottleneck_width
else:
stage_kargs["block_class"] = BottleneckBlock
blocks = (make_trident_stage(**stage_kargs) if stage_idx
== trident_stage_idx else make_stage(**stage_kargs))
in_channels = out_channels
out_channels *= 2
bottleneck_channels *= 2
if freeze_at >= stage_idx:
for block in blocks:
block.freeze()
stages.append(blocks)
return ResNet(stem, stages, out_features=out_features)
def build_resnet_backbone(cfg, input_shape):
"""
Create a ResNet instance from config.
Returns:
ResNet: a :class:`ResNet` instance.
"""
depth = cfg.MODEL.RESNETS.DEPTH
stem_width = {
18: 32,
34: 32,
50: 32,
101: 64,
152: 64,
200: 64,
269: 64
}[depth]
radix = cfg.MODEL.RESNETS.RADIX
deep_stem = cfg.MODEL.RESNETS.DEEP_STEM or (radix > 1)
# need registration of new blocks/stems?
norm = cfg.MODEL.RESNETS.NORM
activation = cfg.MODEL.RESNETS.ACTIVATION
stem = BasicStem(
in_channels=input_shape.channels,
out_channels=cfg.MODEL.RESNETS.STEM_OUT_CHANNELS,
norm=norm,
activation=activation,
deep_stem=deep_stem,
stem_width=stem_width,
)
freeze_at = cfg.MODEL.BACKBONE.FREEZE_AT
if freeze_at >= 1:
for p in stem.parameters():
p.requires_grad = False
stem = FrozenBatchNorm2d.convert_frozen_batchnorm(stem)
# fmt: off
out_features = cfg.MODEL.RESNETS.OUT_FEATURES
num_groups = cfg.MODEL.RESNETS.NUM_GROUPS
width_per_group = cfg.MODEL.RESNETS.WIDTH_PER_GROUP
bottleneck_channels = num_groups * width_per_group
in_channels = cfg.MODEL.RESNETS.STEM_OUT_CHANNELS
out_channels = cfg.MODEL.RESNETS.RES2_OUT_CHANNELS
stride_in_1x1 = cfg.MODEL.RESNETS.STRIDE_IN_1X1
res5_dilation = cfg.MODEL.RESNETS.RES5_DILATION
deform_on_per_stage = cfg.MODEL.RESNETS.DEFORM_ON_PER_STAGE
deform_modulated = cfg.MODEL.RESNETS.DEFORM_MODULATED
deform_num_groups = cfg.MODEL.RESNETS.DEFORM_NUM_GROUPS
num_classes = cfg.MODEL.RESNETS.NUM_CLASSES
zero_init_residual = cfg.MODEL.RESNETS.ZERO_INIT_RESIDUAL
avd = cfg.MODEL.RESNETS.AVD or (radix > 1)
avg_down = cfg.MODEL.RESNETS.AVG_DOWN or (radix > 1)
bottleneck_width = cfg.MODEL.RESNETS.BOTTLENECK_WIDTH
# fmt: on
assert res5_dilation in {
1, 2
}, "res5_dilation cannot be {}.".format(res5_dilation)
num_blocks_per_stage = {
18: [2, 2, 2, 2],
34: [3, 4, 6, 3],
50: [3, 4, 6, 3],
101: [3, 4, 23, 3],
152: [3, 8, 36, 3],
200: [3, 24, 36, 3],
269: [3, 30, 48, 8],
}[depth]
if depth in [18, 34]:
assert out_channels == 64, "Must set MODEL.RESNETS.RES2_OUT_CHANNELS = 64 for R18/R34"
assert not any(
deform_on_per_stage
), "MODEL.RESNETS.DEFORM_ON_PER_STAGE unsupported for R18/R34"
assert res5_dilation == 1, "Must set MODEL.RESNETS.RES5_DILATION = 1 for R18/R34"
assert num_groups == 1, "Must set MODEL.RESNETS.NUM_GROUPS = 1 for R18/R34"
stages = []
# Avoid creating variables without gradients
# It consumes extra memory and may cause allreduce to fail
out_stage_idx = [{
"res2": 2,
"res3": 3,
"res4": 4,
"res5": 5,
"linear": 5
}[f] for f in out_features]
max_stage_idx = max(out_stage_idx)
in_channels = 2 * stem_width if deep_stem else in_channels
for idx, stage_idx in enumerate(range(2, max_stage_idx + 1)):
dilation = res5_dilation if stage_idx == 5 else 1
first_stride = 1 if idx == 0 or (stage_idx == 5
and dilation == 2) else 2
stage_kargs = {
"num_blocks": num_blocks_per_stage[idx],
"first_stride": first_stride,
"in_channels": in_channels,
"out_channels": out_channels,
"norm": norm,
"activation": activation,
"avd": avd,
"avg_down": avg_down,
"radix": radix,
"bottleneck_width": bottleneck_width,
}
# Use BasicBlock for R18 and R34.
if depth in [18, 34]:
stage_kargs["block_class"] = BasicBlock
else:
stage_kargs["bottleneck_channels"] = bottleneck_channels
stage_kargs["stride_in_1x1"] = stride_in_1x1
stage_kargs["dilation"] = dilation
stage_kargs["num_groups"] = num_groups
if deform_on_per_stage[idx]:
stage_kargs["block_class"] = DeformBottleneckBlock
stage_kargs["deform_modulated"] = deform_modulated
stage_kargs["deform_num_groups"] = deform_num_groups
else:
stage_kargs["block_class"] = BottleneckBlock
blocks = make_stage(**stage_kargs)
in_channels = out_channels
out_channels *= 2
bottleneck_channels *= 2
if freeze_at >= stage_idx:
for block in blocks:
block.freeze()
stages.append(blocks)
return ResNet(stem,
stages,
num_classes=num_classes,
out_features=out_features,
zero_init_residual=zero_init_residual)
def _freeze_module(module):
for p in module.parameters():
p.requires_grad = False
return FrozenBatchNorm2d.convert_frozen_batchnorm(module)
def freeze(self):
for p in self.parameters():
p.requires_grad = False
FrozenBatchNorm2d.convert_frozen_batchnorm(self)
return self
def build_resnet_backbone(cfg, input_shape):
"""
Create a ResNet instance from config.
Returns:
ResNet: a :class:`ResNet` instance.
"""
depth = cfg.MODEL.RESNETS.DEPTH
deep_stem = cfg.MODEL.RESNETS.DEEP_STEM
in_channels = input_shape.channels
out_channels = cfg.MODEL.RESNETS.STEM_OUT_CHANNELS
activation = cfg.MODEL.RESNETS.ACTIVATION
norm = cfg.MODEL.RESNETS.NORM
if deep_stem:
stem_width = {
18: 32,
34: 32,
50: 32,
101: 64,
152: 64,
200: 64,
269: 64
}[depth]
stem = DeepStem(
in_channels,
out_channels,
norm,
activation,
deep_stem=deep_stem,
stem_width=stem_width,
)
else:
assert getattr(cfg.MODEL.RESNETS, "RADIX", 1) <= 1, \
"cfg.MODEL.RESNETS.RADIX: {} > 1".format(cfg.MODEL.RESNETS.RADIX)
stem = BasicStem(in_channels,
out_channels,
norm=norm,
activation=activation)
freeze_at = cfg.MODEL.BACKBONE.FREEZE_AT
if freeze_at >= 1:
for p in stem.parameters():
p.requires_grad = False
stem = FrozenBatchNorm2d.convert_frozen_batchnorm(stem)
# fmt: off
out_features = cfg.MODEL.RESNETS.OUT_FEATURES
num_groups = cfg.MODEL.RESNETS.NUM_GROUPS
width_per_group = cfg.MODEL.RESNETS.WIDTH_PER_GROUP
bottleneck_channels = num_groups * width_per_group
in_channels = cfg.MODEL.RESNETS.STEM_OUT_CHANNELS
out_channels = cfg.MODEL.RESNETS.RES2_OUT_CHANNELS
stride_in_1x1 = cfg.MODEL.RESNETS.STRIDE_IN_1X1
res5_dilation = cfg.MODEL.RESNETS.RES5_DILATION
num_classes = cfg.MODEL.RESNETS.NUM_CLASSES
zero_init_residual = cfg.MODEL.RESNETS.ZERO_INIT_RESIDUAL
# fmt: on
assert res5_dilation in {
1, 2
}, "res5_dilation cannot be {}.".format(res5_dilation)
num_blocks_per_stage = {
18: [2, 2, 2, 2],
34: [3, 4, 6, 3],
50: [3, 4, 6, 3],
101: [3, 4, 23, 3],
152: [3, 8, 36, 3],
200: [3, 24, 36, 3],
269: [3, 30, 48, 8],
}[depth]
# Avoid creating variables without gradients
# which consume extra memory and may cause allreduce to fail
out_stage_idx = [{
"res2": 2,
"res3": 3,
"res4": 4,
"res5": 5,
"linear": 5
}[f] for f in out_features]
max_stage_idx = max(out_stage_idx)
# Apply Deformable Convolution in stages
# Specify if apply deform_conv on Res2, Res3, Res4, Res5
deform_on_per_stage = getattr(cfg.MODEL.RESNETS, "DEFORM_ON_PER_STAGE",
[False] * (max_stage_idx - 1))
if depth in [18, 34]:
assert out_channels == 64, "Must set MODEL.RESNETS.RES2_OUT_CHANNELS = 64 for R18/R34"
assert not any(
deform_on_per_stage
), "MODEL.RESNETS.DEFORM_ON_PER_STAGE unsupported for R18/R34"
assert res5_dilation == 1, "Must set MODEL.RESNETS.RES5_DILATION = 1 for R18/R34"
assert num_groups == 1, "Must set MODEL.RESNETS.NUM_GROUPS = 1 for R18/R34"
stages = []
# See cvpods/configs/base_detection_config.py for details
if not stride_in_1x1 and "torchvision" not in cfg.MODEL.WEIGHTS:
logger.warning(
"Using pretrain weight not from torchvision with MODEL.RESNETS.STRIDE_IN_1X1 == False"
)
elif stride_in_1x1 and "torchvision" in cfg.MODEL.WEIGHTS and cfg.MODEL.WEIGHTS:
logger.warning(
"Using pretrain weight from torchvision with MODEL.RESNETS.STRIDE_IN_1X1 == True"
)
in_channels = 2 * stem_width if deep_stem else in_channels
for idx, stage_idx in enumerate(range(2, max_stage_idx + 1)):
dilation = res5_dilation if stage_idx == 5 else 1
first_stride = 1 if idx == 0 or (stage_idx == 5
and dilation == 2) else 2
stage_kargs = {
"num_blocks": num_blocks_per_stage[idx],
"first_stride": first_stride,
"in_channels": in_channels,
"out_channels": out_channels,
"norm": norm,
"activation": activation,
}
# Use BasicBlock for R18 and R34.
if depth in [18, 34]:
stage_kargs["block_class"] = BasicBlock
else:
stage_kargs["bottleneck_channels"] = bottleneck_channels
stage_kargs["stride_in_1x1"] = stride_in_1x1
stage_kargs["dilation"] = dilation
stage_kargs["num_groups"] = num_groups
if deform_on_per_stage[idx]:
stage_kargs["block_class"] = DeformBottleneckBlock
# Use True to use modulated deform_conv (DeformableV2);
# Use False for DeformableV1.
stage_kargs[
"deform_modulated"] = cfg.MODEL.RESNETS.DEFORM_MODULATED
# Number of groups in deformable conv.
stage_kargs[
"deform_num_groups"] = cfg.MODEL.RESNETS.DEFORM_NUM_GROUPS
elif hasattr(cfg.MODEL.RESNETS, "RADIX"):
# Radix in ResNeSt
radix = cfg.MODEL.RESNETS.RADIX
# Apply avg after conv2 in the BottleBlock
# When AVD=True, the STRIDE_IN_1X1 should be False
avd = cfg.MODEL.RESNETS.AVD or (radix > 1)
# Apply avg_down to the downsampling layer for residual path
avg_down = cfg.MODEL.RESNETS.AVG_DOWN or (radix > 1)
# Bottleneck_width in ResNeSt
bottleneck_width = cfg.MODEL.RESNETS.BOTTLENECK_WIDTH
stage_kargs["block_class"] = AVDBottleneckBlock
stage_kargs["avd"] = avd
stage_kargs["avg_down"] = avg_down
stage_kargs["radix"] = radix
stage_kargs["bottleneck_width"] = bottleneck_width
else:
stage_kargs["block_class"] = BottleneckBlock
blocks = make_stage(**stage_kargs)
in_channels = out_channels
out_channels *= 2
bottleneck_channels *= 2
if freeze_at >= stage_idx:
for block in blocks:
block.freeze()
stages.append(blocks)
return ResNet(stem,
stages,
num_classes=num_classes,
out_features=out_features,
zero_init_residual=zero_init_residual)