mirrors/mmsegmentation
1
0
Fork 0
mirror of https://github.com/open-mmlab/mmsegmentation.git synced 2025-06-03 22:03:48 +08:00
Code Issues Projects Releases Wiki Activity

Expand_ratio docstrings updated.

Browse Source
This commit is contained in:
johnzja 2020-08-14 13:13:24 +08:00
parent 88a123d16f
commit a9ac0d8188
2 changed files with 15 additions and 17 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

2
configs/_base_/models/fast_scnn.py
View File

@ -7,7 +7,7 @@ model = dict(
downsample_dw_channels=(32, 48),
global_in_channels=64,
global_block_channels=(64, 96, 128),
global_block_downsample_factors=(2, 2, 1),
global_block_strides=(2, 2, 1),
global_out_channels=128,
higher_in_channels=64,
lower_in_channels=128,

30
mmseg/models/backbones/fast_scnn.py
View File

@ -79,13 +79,14 @@ class GlobalFeatureExtractor(nn.Module):
Default: (64, 96, 128)
out_channels(int): Number of output channels of the GFE module.
Default: 128
expand_ratio (int): Upsampling factor of each Inverted Residual
module. Default: 6
expand_ratio (int): Adjusts number of channels of the hidden layer
in InvertedResidual by this amount.
Default: 6
num_blocks (tuple[int]): Tuple of ints. Each int specifies the
number of times each Inverted Residual module is repeated.
The repeated Inverted Residual modules are called a 'group'.
Default: (3, 3, 3)
downsample_factors (tuple[int]): Tuple of ints. Each int specifies
strides (tuple[int]): Tuple of ints. Each int specifies
the downsampling factor of each 'group'.
Default: (2, 2, 1)
pool_scales (tuple[int]): Tuple of ints. Each int specifies
@ -106,7 +107,7 @@ class GlobalFeatureExtractor(nn.Module):
out_channels=128,
expand_ratio=6,
num_blocks=(3, 3, 3),
downsample_factors=(2, 2, 1),
strides=(2, 2, 1),
pool_scales=(1, 2, 3, 6),
conv_cfg=None,
norm_cfg=dict(type='BN'),
@ -118,17 +119,14 @@ class GlobalFeatureExtractor(nn.Module):
self.act_cfg = act_cfg
assert len(block_channels) == len(num_blocks) == 3
self.bottleneck1 = self._make_layer(in_channels, block_channels[0],
num_blocks[0],
downsample_factors[0],
num_blocks[0], strides[0],
expand_ratio)
self.bottleneck2 = self._make_layer(block_channels[0],
block_channels[1], num_blocks[1],
downsample_factors[1],
expand_ratio)
strides[1], expand_ratio)
self.bottleneck3 = self._make_layer(block_channels[1],
block_channels[2], num_blocks[2],
downsample_factors[2],
expand_ratio)
strides[2], expand_ratio)
self.ppm = PPM(
pool_scales,
block_channels[2],
@ -269,8 +267,8 @@ class FastSCNN(nn.Module):
the output channels for each of the MobileNet-v2 bottleneck
residual blocks in GFE.
Default: (64, 96, 128).
global_block_downsample_factors (tuple[int]): Tuple of integers
that describe the downsampling factors for each of the
global_block_strides (tuple[int]): Tuple of integers
that describe the strides (downsampling factors) for each of the
MobileNet-v2 bottleneck residual blocks in GFE.
Default: (2, 2, 1).
global_out_channels (int): Number of output channels of GFE.
@ -303,7 +301,7 @@ class FastSCNN(nn.Module):
downsample_dw_channels=(32, 48),
global_in_channels=64,
global_block_channels=(64, 96, 128),
global_block_downsample_factors=(2, 2, 1),
global_block_strides=(2, 2, 1),
global_out_channels=128,
higher_in_channels=64,
lower_in_channels=128,
@ -324,7 +322,7 @@ class FastSCNN(nn.Module):
# Calculate scale factor used in FFM.
self.scale_factor = 1
for factor in global_block_downsample_factors:
for factor in global_block_strides:
self.scale_factor *= factor
self.in_channels = in_channels
@ -332,7 +330,7 @@ class FastSCNN(nn.Module):
self.downsample_dw_channels2 = downsample_dw_channels[1]
self.global_in_channels = global_in_channels
self.global_block_channels = global_block_channels
self.global_block_downsample_factors = global_block_downsample_factors
self.global_block_strides = global_block_strides
self.global_out_channels = global_out_channels
self.higher_in_channels = higher_in_channels
self.lower_in_channels = lower_in_channels
@ -353,7 +351,7 @@ class FastSCNN(nn.Module):
global_in_channels,
global_block_channels,
global_out_channels,
downsample_factors=self.global_block_downsample_factors,
downsample_factors=self.global_block_strides,
conv_cfg=self.conv_cfg,
norm_cfg=self.norm_cfg,
act_cfg=self.act_cfg,