-
Notifications
You must be signed in to change notification settings - Fork 8
/
se_block.py
22 lines (18 loc) · 867 Bytes
/
se_block.py
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
import torch
import torch.nn as nn
import torch.nn.functional as F
# https://openaccess.thecvf.com/content_cvpr_2018/html/Hu_Squeeze-and-Excitation_Networks_CVPR_2018_paper.html
class SEBlock(nn.Module):
def __init__(self, input_channels, internal_neurons):
super(SEBlock, self).__init__()
self.down = nn.Conv2d(in_channels=input_channels, out_channels=internal_neurons, kernel_size=1, stride=1, bias=True)
self.up = nn.Conv2d(in_channels=internal_neurons, out_channels=input_channels, kernel_size=1, stride=1, bias=True)
self.input_channels = input_channels
def forward(self, inputs):
x = F.avg_pool2d(inputs, kernel_size=inputs.size(3))
x = self.down(x)
x = F.relu(x)
x = self.up(x)
x = torch.sigmoid(x)
x = x.view(-1, self.input_channels, 1, 1)
return inputs * x