1 研究思路
嘗試提出一個和架構無關的新型CNN結構來提升CNN的性能。穩重提取了非對稱卷積塊(ACB),使用三個並行的代替原始的卷積核進行特徵提取。
2 結構
2.1 非對稱卷積
非對稱卷積基本上是相對對稱卷積來說的,對稱卷積一般爲,非對稱卷積一般爲或者兩類。在inceptionv3中基本上證明了垂直和水平兩個方向的非對稱卷積並行鏈接某種程度上等效於單個對稱卷積且更少的參數,並且vgg證明了多個小卷積核串聯等價於單個大卷積核且更少的參數。同理多個非對稱卷積的串聯可以擁有更廣的感受野。
2.2 公式推導
對於一般的卷積,有:
- 表示輸入;
- 表示卷積核;
- 表示輸出特徵圖;
- 表示卷積操作。
對以上輸出經過bn的結果爲:
- 表示均值;
- 表示標準差;
- 表示縮放係數;
- 表示偏移量。
卷積可加性:
- 爲兩個兼容尺寸的2d核;
- 爲輸入矩陣;
- 爲按位置求和。
2.3 code
class CropLayer(nn.Module):
# E.g., (-1, 0) means this layer should crop the first and last rows of the feature map. And (0, -1) crops the first and last columns
def __init__(self, crop_set):
super(CropLayer, self).__init__()
self.rows_to_crop = - crop_set[0]
self.cols_to_crop = - crop_set[1]
assert self.rows_to_crop >= 0
assert self.cols_to_crop >= 0
def forward(self, input):
return input[:, :, self.rows_to_crop:-self.rows_to_crop, self.cols_to_crop:-self.cols_to_crop]
class ACBlock(nn.Module):
def __init__(self, in_channels, out_channels, kernel_size, stride=1, padding=0, dilation=1, groups=1, padding_mode='zeros', deploy=False):
super(ACBlock, self).__init__()
self.deploy = deploy
if deploy:
self.fused_conv = nn.Conv2d(in_channels=in_channels, out_channels=out_channels, kernel_size=(kernel_size,kernel_size), stride=stride,
padding=padding, dilation=dilation, groups=groups, bias=True, padding_mode=padding_mode)
else:
self.square_conv = nn.Conv2d(in_channels=in_channels, out_channels=out_channels,
kernel_size=(kernel_size, kernel_size), stride=stride,
padding=padding, dilation=dilation, groups=groups, bias=False,
padding_mode=padding_mode)
self.square_bn = nn.BatchNorm2d(num_features=out_channels)
center_offset_from_origin_border = padding - kernel_size // 2
ver_pad_or_crop = (center_offset_from_origin_border + 1, center_offset_from_origin_border)
hor_pad_or_crop = (center_offset_from_origin_border, center_offset_from_origin_border + 1)
if center_offset_from_origin_border >= 0:
self.ver_conv_crop_layer = nn.Identity()
ver_conv_padding = ver_pad_or_crop
self.hor_conv_crop_layer = nn.Identity()
hor_conv_padding = hor_pad_or_crop
else:
self.ver_conv_crop_layer = CropLayer(crop_set=ver_pad_or_crop)
ver_conv_padding = (0, 0)
self.hor_conv_crop_layer = CropLayer(crop_set=hor_pad_or_crop)
hor_conv_padding = (0, 0)
self.ver_conv = nn.Conv2d(in_channels=in_channels, out_channels=out_channels, kernel_size=(3, 1),
stride=stride,
padding=ver_conv_padding, dilation=dilation, groups=groups, bias=False,
padding_mode=padding_mode)
self.hor_conv = nn.Conv2d(in_channels=in_channels, out_channels=out_channels, kernel_size=(1, 3),
stride=stride,
padding=hor_conv_padding, dilation=dilation, groups=groups, bias=False,
padding_mode=padding_mode)
self.ver_bn = nn.BatchNorm2d(num_features=out_channels)
self.hor_bn = nn.BatchNorm2d(num_features=out_channels)
def forward(self, input):
if self.deploy:
return self.fused_conv(input)
else:
square_outputs = self.square_conv(input)
square_outputs = self.square_bn(square_outputs)
# print(square_outputs.size())
# return square_outputs
vertical_outputs = self.ver_conv_crop_layer(input)
vertical_outputs = self.ver_conv(vertical_outputs)
vertical_outputs = self.ver_bn(vertical_outputs)
# print(vertical_outputs.size())
horizontal_outputs = self.hor_conv_crop_layer(input)
horizontal_outputs = self.hor_conv(horizontal_outputs)
horizontal_outputs = self.hor_bn(horizontal_outputs)
# print(horizontal_outputs.size())
return square_outputs + vertical_outputs + horizontal_outputs
3 結果
消融實驗: