def create_module(self) -> 'CenternetDetectionHead':
return CenternetDetectionHead(
conv_in_channels=self.conv_in_channels,
conv_out_channels=self.conv_out_channels,
categories_count=self.categories_count,
base_params=BaseModelParams(name=self.name),
config=self)
def dla34(pretrained=False, **kwargs):
model = DLA(levels=[1, 1, 1, 2, 2, 1],
channels=[16, 32, 64, 128, 256, 512],
block=BasicBlock,
params=BaseModelParams(name='dla34'),
**kwargs)
if pretrained:
model.load_pretrained_model(data='imagenet',
name='dla34',
hash='ba72cf86')
return model
def dla60x(pretrained=False, **kwargs):
BottleneckX.expansion = 2
model = DLA(levels=[1, 1, 1, 2, 3, 1],
channels=[16, 32, 128, 256, 512, 1024],
params=BaseModelParams(name='dla60x'),
block=BottleneckX,
**kwargs)
if pretrained:
model.load_pretrained_model(data='imagenet',
name='dla60x',
hash='d15cacda')
return model
def dla60x_c(pretrained=False, **kwargs):
BottleneckX.expansion = 2
model = DLA(levels=[1, 1, 1, 2, 3, 1],
channels=[16, 32, 64, 64, 128, 256],
block=BottleneckX,
params=BaseModelParams(name='dla60x_c'),
**kwargs)
if pretrained:
model.load_pretrained_model(data='imagenet',
name='dla60x_c',
hash='b870c45c')
return model
def dla169(pretrained=False, **kwargs):
Bottleneck.expansion = 2
model = DLA(levels=[1, 1, 2, 3, 5, 1],
channels=[16, 32, 128, 256, 512, 1024],
block=Bottleneck,
residual_root=True,
params=BaseModelParams(name='dla169'),
**kwargs)
if pretrained:
model.load_pretrained_model(data='imagenet',
name='dla169',
hash='0914e092')
return model
def dla102x2(pretrained=False, **kwargs):
BottleneckX.cardinality = 64
model = DLA(levels=[1, 1, 1, 3, 4, 1],
channels=[16, 32, 128, 256, 512, 1024],
block=BottleneckX,
residual_root=True,
params=BaseModelParams(name='dla102x2'),
**kwargs)
if pretrained:
model.load_pretrained_model(data='imagenet',
name='dla102x2',
hash='262837b6')
return model
def dla102x(pretrained=False, **kwargs):
BottleneckX.expansion = 2
model = DLA(levels=[1, 1, 1, 3, 4, 1],
channels=[16, 32, 128, 256, 512, 1024],
block=BottleneckX,
residual_root=True,
params=BaseModelParams(name='dla102x'),
**kwargs)
if pretrained:
model.load_pretrained_model(data='imagenet',
name='dla102x',
hash='ad62be81')
return model