-
Notifications
You must be signed in to change notification settings - Fork 0
/
cnn2.py
32 lines (30 loc) · 1.4 KB
/
cnn2.py
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
import torch.nn.functional as F
import torch.nn as nn
import torch
class cnn2_zjc(nn.Module):#batchsize,16,16,16
def __init__(self) :
super(cnn2_zjc,self).__init__()
self.cnn2_part0= nn.Sequential(nn.Conv2d(in_channels=16,out_channels=8,kernel_size=3,stride=1,padding=1),
nn.BatchNorm2d(8),
nn.ReLU()
)
self.cnn2_part1= nn.Sequential(nn.Conv2d(in_channels=8,out_channels=3,kernel_size=3,stride=1,padding=1),
nn.BatchNorm2d(3),
nn.ReLU()
)
self.cnn2_part2= nn.Sequential(nn.Conv2d(in_channels=3,out_channels=3,kernel_size=3,stride=1,padding=1),
nn.BatchNorm2d(3),
nn.ReLU()
)
self.cnn2_part3= nn.Sequential(nn.Conv2d(in_channels=3,out_channels=3,kernel_size=3,stride=1,padding=1),
nn.BatchNorm2d(3)
)
def forward(self,x):
x=self.cnn2_part0(x)
x=F.interpolate(input=x,size=(32,32),mode='bilinear')
x=self.cnn2_part1(x)
x=F.interpolate(input=x,size=(64,64),mode='bilinear')
x=self.cnn2_part2(x)
x=F.interpolate(input=x,size=(128,128),mode='bilinear')
x=self.cnn2_part3(x)
return torch.sigmoid(x)