Unet网络结构
extra_large的Unet网络结构:模型大小:1944756KB,1.85GBimport torch.nn as nn
import torch
from torch import autograd
import torchvision.models as models
# 导入模型结构
#resnet50 = models.resnet50(pretrained=True)
# 加载预先下载好的预训练参数到resnet18
#resnet50.load_state_dict(torch.load('resnet50-5c106cde.pth'))
class DoubleConv(nn.Module):
def __init__(self, in_ch, out_ch):
super(DoubleConv, self).__init__()
self.conv = nn.Sequential(
nn.Conv2d(in_ch, out_ch, 3, padding=1),
nn.BatchNorm2d(out_ch),
nn.ReLU(inplace=True),
nn.Conv2d(out_ch, out_ch, 3, padding=1),
nn.BatchNorm2d(out_ch),
nn.ReLU(inplace=True)
)
def forward(self, input):
return self.conv(input)
class Unet(nn.Module):
def __init__(self,in_ch,out_ch):
super(Unet, self).__init__()
self.conv1 = DoubleConv(in_ch, 16)
self.pool1 = nn.MaxPool2d(2)
self.conv2 = DoubleConv(16, 32)
self.pool2 = nn.MaxPool2d(2)
self.conv3 = DoubleConv(32, 64)
self.pool3 = nn.MaxPool2d(2)
self.conv4 = DoubleConv(64, 128)
self.pool4 = nn.MaxPool2d(2)
self.conv5 = DoubleConv(128, 256)
self.pool5 = nn.MaxPool2d(2)
self.conv6 = DoubleConv(256, 512)
self.pool6 = nn.MaxPool2d(2)
self.conv7 = DoubleConv(512, 1024)
self.pool7 = nn.MaxPool2d(2)
self.conv8 = DoubleConv(1024, 2048)
self.pool8 = nn.MaxPool2d(2)
self.conv9 = DoubleConv(2048, 4096)
self.up10 = nn.ConvTranspose2d(4096, 2048, 2, stride=2)
self.conv10 = DoubleConv(4096, 2048)
self.up11 = nn.ConvTranspose2d(2048, 1024, 2, stride=2)
self.conv11 = DoubleConv(2048, 1024)
self.up12 = nn.ConvTranspose2d(1024, 512, 2, stride=2)
self.conv12 = DoubleConv(1024, 512)
self.up13 = nn.ConvTranspose2d(512, 256, 2, stride=2)
self.conv13 = DoubleConv(512, 256)
self.up14 = nn.ConvTranspose2d(256, 128, 2, stride=2)
self.conv14 = DoubleConv(256, 128)
self.up15 = nn.ConvTranspose2d(128, 64, 2, stride=2)
self.conv15 = DoubleConv(128, 64)
self.up16 = nn.ConvTranspose2d(64, 32, 2, stride=2)
self.conv16 = DoubleConv(64, 32)
self.up17 = nn.ConvTranspose2d(32, 16, 2, stride=2)
self.conv17 = DoubleConv(32, 16)
self.conv18 = nn.Conv2d(16,out_ch, 1)
def forward(self,x):
c1=self.conv1(x)
p1=self.pool1(c1)
c2=self.conv2(p1)
p2=self.pool2(c2)
c3=self.conv3(p2)
p3=self.pool3(c3)
c4=self.conv4(p3)
p4=self.pool4(c4)
c5=self.conv5(p4)
p5=self.pool5(c5)
c6=self.conv6(p5)
p6=self.pool6(c6)
c7=self.conv7(p6)
p7=self.pool7(c7)
c8=self.conv8(p7)
p8=self.pool8(c8)
c9=self.conv9(p8)
up_10= self.up10(c9)
merge10 = torch.cat(, dim=1)
c10=self.conv10(merge10)
up_11= self.up11(c10)
merge11 = torch.cat(, dim=1)
c11=self.conv11(merge11)
up_12= self.up12(c11)
merge12 = torch.cat(, dim=1)
c12=self.conv12(merge12)
up_13=self.up13(c12)
merge13 = torch.cat(, dim=1)
c13=self.conv13(merge13)
up_14=self.up14(c13)
merge14 = torch.cat(, dim=1)
c14=self.conv14(merge14)
up_15=self.up15(c14)
merge15 = torch.cat(,dim=1)
c15=self.conv15(merge15)
up_16=self.up16(c15)
merge16 = torch.cat(,dim=1)
c16=self.conv16(merge16)
up_17=self.up17(c16)
merge17 = torch.cat(,dim=1)
c17=self.conv17(merge17)
c18=self.conv18(c17)
out = nn.Sigmoid()(c18)
return out
large Unet网络1:模型大小:121652KB,118MB
import torch.nn as nn
import torch
from torch import autograd
import torchvision.models as models
# 导入模型结构
#resnet50 = models.resnet50(pretrained=True)
# 加载预先下载好的预训练参数到resnet18
#resnet50.load_state_dict(torch.load('resnet50-5c106cde.pth'))
class DoubleConv(nn.Module):
def __init__(self, in_ch, out_ch):
super(DoubleConv, self).__init__()
self.conv = nn.Sequential(
nn.Conv2d(in_ch, out_ch, 3, padding=1),
nn.BatchNorm2d(out_ch),
nn.ReLU(inplace=True),
nn.Conv2d(out_ch, out_ch, 3, padding=1),
nn.BatchNorm2d(out_ch),
nn.ReLU(inplace=True)
)
def forward(self, input):
return self.conv(input)
class Unet(nn.Module):
def __init__(self,in_ch,out_ch):
super(Unet, self).__init__()
self.conv1 = DoubleConv(in_ch, 16)
self.pool1 = nn.MaxPool2d(2)
self.conv2 = DoubleConv(16, 32)
self.pool2 = nn.MaxPool2d(2)
self.conv3 = DoubleConv(32, 64)
self.pool3 = nn.MaxPool2d(2)
self.conv4 = DoubleConv(64, 128)
self.pool4 = nn.MaxPool2d(2)
self.conv5 = DoubleConv(128, 256)
self.pool5 = nn.MaxPool2d(2)
self.conv6 = DoubleConv(256, 512)
self.pool6 = nn.MaxPool2d(2)
self.conv7 = DoubleConv(512, 1024)
self.up8 = nn.ConvTranspose2d(1024, 512, 2, stride=2)
self.conv8 = DoubleConv(1024, 512)
self.up9 = nn.ConvTranspose2d(512, 256, 2, stride=2)
self.conv9 = DoubleConv(512, 256)
self.up10 = nn.ConvTranspose2d(256, 128, 2, stride=2)
self.conv10 = DoubleConv(256, 128)
self.up11 = nn.ConvTranspose2d(128, 64, 2, stride=2)
self.conv11 = DoubleConv(128, 64)
self.up12 = nn.ConvTranspose2d(64, 32, 2, stride=2)
self.conv12 = DoubleConv(64, 32)
self.up13 = nn.ConvTranspose2d(32, 16, 2, stride=2)
self.conv13 = DoubleConv(32, 16)
self.conv14 = nn.Conv2d(16,out_ch, 1)
def forward(self,x):
c1=self.conv1(x)
p1=self.pool1(c1)
c2=self.conv2(p1)
p2=self.pool2(c2)
c3=self.conv3(p2)
p3=self.pool3(c3)
c4=self.conv4(p3)
p4=self.pool4(c4)
c5=self.conv5(p4)
p5=self.pool5(c5)
c6=self.conv6(p5)
p6=self.pool6(c6)
c7=self.conv7(p6)
up_8= self.up8(c7)
merge8 = torch.cat(, dim=1)
c8=self.conv8(merge8)
up_9=self.up9(c8)
merge9 = torch.cat(, dim=1)
c9=self.conv9(merge9)
up_10=self.up10(c9)
merge10 = torch.cat(, dim=1)
c10=self.conv10(merge10)
up_11=self.up11(c10)
merge11=torch.cat(,dim=1)
c11=self.conv11(merge11)
up_12=self.up12(c11)
merge12=torch.cat(,dim=1)
c12=self.conv12(merge12)
up_13=self.up13(c12)
merge13=torch.cat(,dim=1)
c13=self.conv13(merge13)
c14=self.conv14(c13)
out = nn.Sigmoid()(c14)
return out
large Unet网络2::模型大小:486257KB,474MB
import torch.nn as nn
import torch
from torch import autograd
import torchvision.models as models
# 导入模型结构
#resnet50 = models.resnet50(pretrained=True)
# 加载预先下载好的预训练参数到resnet18
#resnet50.load_state_dict(torch.load('resnet50-5c106cde.pth'))
class DoubleConv(nn.Module):
def __init__(self, in_ch, out_ch):
super(DoubleConv, self).__init__()
self.conv = nn.Sequential(
nn.Conv2d(in_ch, out_ch, 3, padding=1),
nn.BatchNorm2d(out_ch),
nn.ReLU(inplace=True),
nn.Conv2d(out_ch, out_ch, 3, padding=1),
nn.BatchNorm2d(out_ch),
nn.ReLU(inplace=True)
)
def forward(self, input):
return self.conv(input)
class Unet(nn.Module):
def __init__(self,in_ch,out_ch):
super(Unet, self).__init__()
self.conv1 = DoubleConv(in_ch, 32)
self.pool1 = nn.MaxPool2d(2)
self.conv2 = DoubleConv(32, 64)
self.pool2 = nn.MaxPool2d(2)
self.conv3 = DoubleConv(64, 128)
self.pool3 = nn.MaxPool2d(2)
self.conv4 = DoubleConv(128, 256)
self.pool4 = nn.MaxPool2d(2)
self.conv5 = DoubleConv(256, 512)
self.pool5 = nn.MaxPool2d(2)
self.conv6 = DoubleConv(512, 1024)
self.pool6 = nn.MaxPool2d(2)
self.conv7 = DoubleConv(1024, 2048)
self.up8 = nn.ConvTranspose2d(2048, 1024, 2, stride=2)
self.conv8 = DoubleConv(2048, 1024)
self.up9 = nn.ConvTranspose2d(1024, 512, 2, stride=2)
self.conv9 = DoubleConv(1024, 512)
self.up10 = nn.ConvTranspose2d(512, 256, 2, stride=2)
self.conv10 = DoubleConv(512, 256)
self.up11 = nn.ConvTranspose2d(256, 128, 2, stride=2)
self.conv11 = DoubleConv(256, 128)
self.up12 = nn.ConvTranspose2d(128, 64, 2, stride=2)
self.conv12 = DoubleConv(128, 64)
self.up13 = nn.ConvTranspose2d(64, 32, 2, stride=2)
self.conv13 = DoubleConv(64, 32)
self.conv14 = nn.Conv2d(32,out_ch, 1)
def forward(self,x):
c1=self.conv1(x)
p1=self.pool1(c1)
c2=self.conv2(p1)
p2=self.pool2(c2)
c3=self.conv3(p2)
p3=self.pool3(c3)
c4=self.conv4(p3)
p4=self.pool4(c4)
c5=self.conv5(p4)
p5=self.pool5(c5)
c6=self.conv6(p5)
p6=self.pool6(c6)
c7=self.conv7(p6)
up_8= self.up8(c7)
merge8 = torch.cat(, dim=1)
c8=self.conv8(merge8)
up_9=self.up9(c8)
merge9 = torch.cat(, dim=1)
c9=self.conv9(merge9)
up_10=self.up10(c9)
merge10 = torch.cat(, dim=1)
c10=self.conv10(merge10)
up_11=self.up11(c10)
merge11=torch.cat(,dim=1)
c11=self.conv11(merge11)
up_12=self.up12(c11)
merge12=torch.cat(,dim=1)
c12=self.conv12(merge12)
up_13=self.up13(c12)
merge13=torch.cat(,dim=1)
c13=self.conv13(merge13)
c14=self.conv14(c13)
out = nn.Sigmoid()(c14)
return out
large Unet网络::模型大小:121329KB,118MB
import torch.nn as nn
import torch
from torch import autograd
import torchvision.models as models
# 导入模型结构
#resnet50 = models.resnet50(pretrained=True)
# 加载预先下载好的预训练参数到resnet18
#resnet50.load_state_dict(torch.load('resnet50-5c106cde.pth'))
class DoubleConv(nn.Module):
def __init__(self, in_ch, out_ch):
super(DoubleConv, self).__init__()
self.conv = nn.Sequential(
nn.Conv2d(in_ch, out_ch, 3, padding=1),
nn.BatchNorm2d(out_ch),
nn.ReLU(inplace=True),
nn.Conv2d(out_ch, out_ch, 3, padding=1),
nn.BatchNorm2d(out_ch),
nn.ReLU(inplace=True)
)
def forward(self, input):
return self.conv(input)
class Unet(nn.Module):
def __init__(self,in_ch,out_ch):
super(Unet, self).__init__()
self.conv1 = DoubleConv(in_ch, 64)
self.pool1 = nn.MaxPool2d(2)
self.conv2 = DoubleConv(64, 128)
self.pool2 = nn.MaxPool2d(2)
self.conv3 = DoubleConv(128, 256)
self.pool3 = nn.MaxPool2d(2)
self.conv4 = DoubleConv(256, 512)
self.pool4 = nn.MaxPool2d(2)
self.conv5 = DoubleConv(512, 1024)
self.up6 = nn.ConvTranspose2d(1024, 512, 2, stride=2)
self.conv6 = DoubleConv(1024, 512)
self.up7 = nn.ConvTranspose2d(512, 256, 2, stride=2)
self.conv7 = DoubleConv(512, 256)
self.up8 = nn.ConvTranspose2d(256, 128, 2, stride=2)
self.conv8 = DoubleConv(256, 128)
self.up9 = nn.ConvTranspose2d(128, 64, 2, stride=2)
self.conv9 = DoubleConv(128, 64)
self.conv10 = nn.Conv2d(64,out_ch, 1)
def forward(self,x):
c1=self.conv1(x)
p1=self.pool1(c1)
c2=self.conv2(p1)
p2=self.pool2(c2)
c3=self.conv3(p2)
p3=self.pool3(c3)
c4=self.conv4(p3)
p4=self.pool4(c4)
c5=self.conv5(p4)
up_6= self.up6(c5)
merge6 = torch.cat(, dim=1)
c6=self.conv6(merge6)
up_7=self.up7(c6)
merge7 = torch.cat(, dim=1)
c7=self.conv7(merge7)
up_8=self.up8(c7)
merge8 = torch.cat(, dim=1)
c8=self.conv8(merge8)
up_9=self.up9(c8)
merge9=torch.cat(,dim=1)
c9=self.conv9(merge9)
c10=self.conv10(c9)
out = nn.Sigmoid()(c10)
return out
normal网络::模型大小:30126KB,29.4MB
import torch.nn as nn
import torch
from torch import autograd
import torchvision.models as models
# 导入模型结构
#resnet50 = models.resnet50(pretrained=True)
# 加载预先下载好的预训练参数到resnet18
#resnet50.load_state_dict(torch.load('resnet50-5c106cde.pth'))
class DoubleConv(nn.Module):
def __init__(self, in_ch, out_ch):
super(DoubleConv, self).__init__()
self.conv = nn.Sequential(
nn.Conv2d(in_ch, out_ch, 3, padding=1),
nn.BatchNorm2d(out_ch),
nn.ReLU(inplace=True),
nn.Conv2d(out_ch, out_ch, 3, padding=1),
nn.BatchNorm2d(out_ch),
nn.ReLU(inplace=True)
)
def forward(self, input):
return self.conv(input)
class Unet(nn.Module):
def __init__(self,in_ch,out_ch):
super(Unet, self).__init__()
self.conv1 = DoubleConv(in_ch, 64)
self.pool1 = nn.MaxPool2d(2)
self.conv2 = DoubleConv(64, 128)
self.pool2 = nn.MaxPool2d(2)
self.conv3 = DoubleConv(128, 256)
self.pool3 = nn.MaxPool2d(2)
self.conv4 = DoubleConv(256, 512)
self.up5 = nn.ConvTranspose2d(512, 256, 2, stride=2)
self.conv5 = DoubleConv(512, 256)
self.up6 = nn.ConvTranspose2d(256, 128, 2, stride=2)
self.conv6 = DoubleConv(256, 128)
self.up7 = nn.ConvTranspose2d(128, 64, 2, stride=2)
self.conv7 = DoubleConv(128, 64)
self.conv8 = nn.Conv2d(64,out_ch, 1)
def forward(self,x):
c1=self.conv1(x)
p1=self.pool1(c1)
c2=self.conv2(p1)
p2=self.pool2(c2)
c3=self.conv3(p2)
p3=self.pool3(c3)
c4=self.conv4(p3)
up_5= self.up5(c4)
merge5 = torch.cat(, dim=1)
c5=self.conv5(merge5)
up_6=self.up6(c5)
merge6 = torch.cat(, dim=1)
c6=self.conv6(merge6)
up_7=self.up7(c6)
merge7 = torch.cat(,dim=1)
c7=self.conv7(merge7)
c8=self.conv8(c7)
out = nn.Sigmoid()(c8)
return out
normal 网络2::模型大小:7306KB,7.13MB
import torch.nn as nn
import torch
from torch import autograd
import torchvision.models as models
# 导入模型结构
#resnet50 = models.resnet50(pretrained=True)
# 加载预先下载好的预训练参数到resnet18
#resnet50.load_state_dict(torch.load('resnet50-5c106cde.pth'))
class DoubleConv(nn.Module):
def __init__(self, in_ch, out_ch):
super(DoubleConv, self).__init__()
self.conv = nn.Sequential(
nn.Conv2d(in_ch, out_ch, 3, padding=1),
nn.BatchNorm2d(out_ch),
nn.ReLU(inplace=True),
nn.Conv2d(out_ch, out_ch, 3, padding=1),
nn.BatchNorm2d(out_ch),
nn.ReLU(inplace=True)
)
def forward(self, input):
return self.conv(input)
class Unet(nn.Module):
def __init__(self,in_ch,out_ch):
super(Unet, self).__init__()
# 128x128
self.conv1 = DoubleConv(in_ch, 64)
self.pool1 = nn.MaxPool2d(2)
self.conv2 = DoubleConv(64, 128)
self.pool2 = nn.MaxPool2d(2)
self.conv3 = DoubleConv(128, 256)
self.up4 = nn.ConvTranspose2d(256, 128, 2, stride=2)
self.conv4 = DoubleConv(256, 128)
self.up5 = nn.ConvTranspose2d(128, 64, 2, stride=2)
self.conv5 = DoubleConv(128, 64)
self.conv6 = nn.Conv2d(64,out_ch, 1)
def forward(self,x):
c1=self.conv1(x)
p1=self.pool1(c1)
c2=self.conv2(p1)
p2=self.pool2(c2)
c3=self.conv3(p2)
up_4=self.up4(c3)
merge4 = torch.cat(, dim=1)
c4=self.conv4(merge4)
up_5=self.up5(c4)
merge5 = torch.cat(,dim=1)
c5=self.conv5(merge5)
c6=self.conv6(c5)
out = nn.Sigmoid()(c6)
return out normal 网络3::模型大小:29126KB,28.4MB
import torch.nn as nn
import torch
from torch import autograd
import torchvision.models as models
# 导入模型结构
#resnet50 = models.resnet50(pretrained=True)
# 加载预先下载好的预训练参数到resnet18
#resnet50.load_state_dict(torch.load('resnet50-5c106cde.pth'))
class DoubleConv(nn.Module):
def __init__(self, in_ch, out_ch):
super(DoubleConv, self).__init__()
self.conv = nn.Sequential(
nn.Conv2d(in_ch, out_ch, 3, padding=1),
nn.BatchNorm2d(out_ch),
nn.ReLU(inplace=True),
nn.Conv2d(out_ch, out_ch, 3, padding=1),
nn.BatchNorm2d(out_ch),
nn.ReLU(inplace=True)
)
def forward(self, input):
return self.conv(input)
class Unet(nn.Module):
def __init__(self,in_ch,out_ch):
super(Unet, self).__init__()
# 256x256
self.conv1 = DoubleConv(in_ch, 128)
self.pool1 = nn.MaxPool2d(2)
self.conv2 = DoubleConv(128, 256)
self.pool2 = nn.MaxPool2d(2)
self.conv3 = DoubleConv(256, 512)
self.up4 = nn.ConvTranspose2d(512, 256, 2, stride=2)
self.conv4 = DoubleConv(512, 256)
self.up5 = nn.ConvTranspose2d(256, 128, 2, stride=2)
self.conv5 = DoubleConv(256, 128)
self.conv6 = nn.Conv2d(128,out_ch, 1)
def forward(self,x):
c1=self.conv1(x)
p1=self.pool1(c1)
c2=self.conv2(p1)
p2=self.pool2(c2)
c3=self.conv3(p2)
up_4=self.up4(c3)
merge4 = torch.cat(, dim=1)
c4=self.conv4(merge4)
up_5=self.up5(c4)
merge5 = torch.cat(,dim=1)
c5=self.conv5(merge5)
c6=self.conv6(c5)
out = nn.Sigmoid()(c6)
return out
normal 网络4::模型大小:120329KB,117MB
# -*- coding: utf-8 -*-
"""
Created on Sat Jun 22 12:05:39 2019
@author: Solem
"""
import torch.nn as nn
import torch
from torch import autograd
import torchvision.models as models
# 导入模型结构
#resnet50 = models.resnet50(pretrained=True)
# 加载预先下载好的预训练参数到resnet18
#resnet50.load_state_dict(torch.load('resnet50-5c106cde.pth'))
class DoubleConv(nn.Module):
def __init__(self, in_ch, out_ch):
super(DoubleConv, self).__init__()
self.conv = nn.Sequential(
nn.Conv2d(in_ch, out_ch, 3, padding=1),
nn.BatchNorm2d(out_ch),
nn.ReLU(inplace=True),
nn.Conv2d(out_ch, out_ch, 3, padding=1),
nn.BatchNorm2d(out_ch),
nn.ReLU(inplace=True)
)
def forward(self, input):
return self.conv(input)
class Unet(nn.Module):
def __init__(self,in_ch,out_ch):
super(Unet, self).__init__()
self.conv1 = DoubleConv(in_ch, 128)
self.pool1 = nn.MaxPool2d(2)
self.conv2 = DoubleConv(128, 256)
self.pool2 = nn.MaxPool2d(2)
self.conv3 = DoubleConv(256, 512)
self.pool3 = nn.MaxPool2d(2)
self.conv4 = DoubleConv(512, 1024)
self.up5 = nn.ConvTranspose2d(1024, 512, 2, stride=2)
self.conv5 = DoubleConv(1024, 512)
self.up6 = nn.ConvTranspose2d(512, 256, 2, stride=2)
self.conv6 = DoubleConv(512, 256)
self.up7 = nn.ConvTranspose2d(256, 128, 2, stride=2)
self.conv7 = DoubleConv(256, 128)
self.conv8 = nn.Conv2d(128,out_ch, 1)
def forward(self,x):
c1=self.conv1(x)
p1=self.pool1(c1)
c2=self.conv2(p1)
p2=self.pool2(c2)
c3=self.conv3(p2)
p3=self.pool3(c3)
c4=self.conv4(p3)
up_5=self.up5(c4)
merge5 = torch.cat(, dim=1)
c5=self.conv5(merge5)
up_6=self.up6(c5)
merge6 = torch.cat(, dim=1)
c6=self.conv6(merge6)
up_7=self.up7(c6)
merge7 = torch.cat(,dim=1)
c7=self.conv7(merge7)
c8=self.conv8(c7)
out = nn.Sigmoid()(c8)
return out
normal 网络5::模型大小:67176KB,66.1MB
import torch.nn as nn
import torch
from torch import autograd
import torchvision.models as models
# 导入模型结构
#resnet50 = models.resnet50(pretrained=True)
# 加载预先下载好的预训练参数到resnet18
#resnet50.load_state_dict(torch.load('resnet50-5c106cde.pth'))
class DoubleConv(nn.Module):
def __init__(self, in_ch, out_ch, kernel_size, stride, padding):
super(DoubleConv, self).__init__()
self.conv = nn.Sequential(
nn.Conv2d(in_ch, out_ch, kernel_size=3, stride=1, padding=1),
nn.BatchNorm2d(out_ch),
nn.ReLU(inplace=True),
nn.Conv2d(out_ch, out_ch, kernel_size=3, stride=1, padding=1),
nn.BatchNorm2d(out_ch),
nn.ReLU(inplace=True)
)
def forward(self, input):
return self.conv(input)
class Unet(nn.Module):
def __init__(self,in_ch,out_ch):
super(Unet, self).__init__()
self.conv1 = DoubleConv(in_ch, 96, kernel_size=3, stride=1, padding=1)
self.pool1 = nn.MaxPool2d(2)
self.conv2 = DoubleConv(96, 192, kernel_size=3, stride=1, padding=1)
self.pool2 = nn.MaxPool2d(2)
self.conv3 = DoubleConv(192, 384, kernel_size=3, stride=1, padding=1)
self.pool3 = nn.MaxPool2d(2)
self.conv4 = DoubleConv(384, 768, kernel_size=3, stride=1, padding=1)
self.up5 = nn.ConvTranspose2d(768, 384, kernel_size=2, stride=2, padding=0)
self.conv5 = DoubleConv(768, 384, kernel_size=3, stride=1, padding=1)
self.up6 = nn.ConvTranspose2d(384, 192, kernel_size=2, stride=2, padding=0)
self.conv6 = DoubleConv(384, 192, kernel_size=3, stride=1, padding=1)
self.up7 = nn.ConvTranspose2d(192, 96, kernel_size=2, stride=2, padding=0)
self.conv7 = DoubleConv(192, 96, kernel_size=3, stride=1, padding=1)
self.conv8 = nn.Conv2d(96, out_ch, 1)
def forward(self,x):
c1=self.conv1(x)
p1=self.pool1(c1)
c2=self.conv2(p1)
p2=self.pool2(c2)
c3=self.conv3(p2)
p3=self.pool3(c3)
c4=self.conv4(p3)
up_5=self.up5(c4)
merge5 = torch.cat(, dim=1)
c5=self.conv5(merge5)
up_6=self.up6(c5)
merge6 = torch.cat(, dim=1)
c6=self.conv6(merge6)
up_7=self.up7(c6)
merge7 = torch.cat(,dim=1)
c7=self.conv7(merge7)
c8=self.conv8(c7)
out = nn.Sigmoid()(c8)
return out
small 网络::模型大小:16400KB,16MB
import torch.nn as nn
import torch
from torch import autograd
import torchvision.models as models
# 导入模型结构
#resnet50 = models.resnet50(pretrained=True)
# 加载预先下载好的预训练参数到resnet18
#resnet50.load_state_dict(torch.load('resnet50-5c106cde.pth'))
class DoubleConv(nn.Module):
def __init__(self, in_ch, out_ch):
super(DoubleConv, self).__init__()
self.conv = nn.Sequential(
nn.Conv2d(in_ch, out_ch, 3, padding=1),
nn.BatchNorm2d(out_ch),
nn.ReLU(inplace=True),
nn.Conv2d(out_ch, out_ch, 3, padding=1),
nn.BatchNorm2d(out_ch),
nn.ReLU(inplace=True)
)
def forward(self, input):
return self.conv(input)
class Unet(nn.Module):
def __init__(self,in_ch,out_ch):
super(Unet, self).__init__()
# 256x256
self.conv1 = DoubleConv(in_ch, 96)
self.pool1 = nn.MaxPool2d(2)
self.conv2 = DoubleConv(96, 192)
self.pool2 = nn.MaxPool2d(2)
self.conv3 = DoubleConv(192, 384)
self.up4 = nn.ConvTranspose2d(384, 192, 2, stride=2)
self.conv4 = DoubleConv(384, 192)
self.up5 = nn.ConvTranspose2d(192, 96, 2, stride=2)
self.conv5 = DoubleConv(192, 96)
self.conv6 = nn.Conv2d(96,out_ch, 1)
def forward(self,x):
c1=self.conv1(x)
p1=self.pool1(c1)
c2=self.conv2(p1)
p2=self.pool2(c2)
c3=self.conv3(p2)
up_4=self.up4(c3)
merge4 = torch.cat(, dim=1)
c4=self.conv4(merge4)
up_5=self.up5(c4)
merge5 = torch.cat(,dim=1)
c5=self.conv5(merge5)
c6=self.conv6(c5)
out = nn.Sigmoid()(c6)
return out
页:
[1]
2