def ResNet_V1_beta_block(scope, base_depth, num_units, stride):
return resnet_v1_utils.Block(scope, Bottleneck, [{
'depth': base_depth * 4,
'depth_bottleneck': base_depth,
'stride': 1,
'unit_rate': 1
}] * (num_units - 1) + [{
'depth': base_depth * 4,
'depth_bottleneck': base_depth,
'stride': stride,
'unit_rate': 1
}])
def ResNetV1Beta_18(num_classes=None,
is_training=None,
global_pool=False,
base_only=False,
min_base_depth=8,
depth_multiplier=1,
root_depth_multiplier=0.25,
output_stride=None,
extract_blocks=None,
input_shape=None,
input_tensor=None,
multi_grid=None,
name='resnet_v1_18'):
if multi_grid is None:
multi_grid = _DEFAULT_MULTI_GRID_RESNET_18
elif len(multi_grid) != 2:
raise ValueError("Expect multi_grid to have length 2.")
depth_func = lambda d: max(int(d * depth_multiplier), min_base_depth)
block4_args = []
for rate in multi_grid:
block4_args.append({
'depth': depth_func(512),
'stride': 1,
'unit_rate': rate
})
blocks = [
ResNet_V1_small_beta_block('block1',
base_depth=depth_func(64),
num_units=2,
stride=2),
ResNet_V1_small_beta_block('block2',
base_depth=depth_func(128),
num_units=2,
stride=2),
ResNet_V1_small_beta_block('block3',
base_depth=depth_func(256),
num_units=2,
stride=2),
resnet_v1_utils.Block('block4', BasicBlock, block4_args)
]
return _ResNetV1Beta(
blocks=blocks,
num_classes=num_classes,
is_training=is_training,
global_pool=global_pool,
base_only=base_only,
root_block=resnet_v1_utils.root_block(root_depth_multiplier),
output_stride=output_stride,
extract_blocks=extract_blocks,
input_shape=input_shape,
input_tensor=input_tensor,
name=name)
def ResNetV1Beta_101(num_classes=None,
is_training=None,
base_only=False,
global_pool=False,
min_base_depth=8,
depth_multiplier=1,
output_stride=None,
extract_blocks=None,
input_shape=None,
input_tensor=None,
multi_grid=None,
name='resnet_v1_101'):
if multi_grid is None:
multi_grid = _DEFAULT_MULTI_GRID
elif len(multi_grid) != 3:
raise ValueError("Expect multi_grid to have length 3.")
depth_func = lambda d: max(int(d * depth_multiplier), min_base_depth)
blocks = [
ResNet_V1_beta_block('block1',
base_depth=depth_func(64),
num_units=3,
stride=2),
ResNet_V1_beta_block('block2',
base_depth=depth_func(128),
num_units=4,
stride=2),
ResNet_V1_beta_block('block3',
base_depth=depth_func(256),
num_units=23,
stride=2),
resnet_v1_utils.Block('block4', Bottleneck,
[{
'depth': 2048,
'depth_bottleneck': depth_func(512),
'stride': 1,
'unit_rate': rate
} for rate in multi_grid])
]
return _ResNetV1Beta(blocks=blocks,
num_classes=num_classes,
is_training=is_training,
base_only=base_only,
global_pool=global_pool,
root_block=resnet_v1_utils.root_block(),
output_stride=output_stride,
extract_blocks=extract_blocks,
input_shape=input_shape,
input_tensor=input_tensor,
name=name)
def ResNet_V1_small_beta_block(scope, base_depth, num_units, stride):
block_args = []
for _ in range(num_units - 1):
block_args.append({'depth': base_depth, 'stride': 1, 'unit_rate': 1})
block_args.append({'depth': base_depth, 'stride': stride, 'unit_rate': 1})
return resnet_v1_utils.Block(scope, BasicBlock, block_args)