在这里,我正在尝试设计一个具有2个卷积层,2个残差单元和一个完全连接的层,然后是一个分类层的CNN模型。我们使用
-3x3卷积内核(resnet单元的步幅为1,卷积层的步幅为2) -ReLU用于激活功能 -max仅在卷积层之后才使用内核2x2和跨度2的池。
我想在隐藏层中定义要素地图的数量为:16、16、16、32、32、32、64(第1层,...,第7层)。
输入->卷积1-> ResNetBlock1->卷积2-> ResNetBlock2-> FC->输出
我在这里写了我的代码。注意:我经常参考本教程。
http://www.pabloruizruiz10.com/resources/CNNs/ResNet-PyTorch.html
import torch
import torch.nn as nn
import torch.nn.functional as F
import torchvision
#3x3 kernel=3
def conv3x3(in_channels, out_channels, stride=1):
return nn.Conv2d(in_channels, out_channels, kernel_size=3, stride=stride, padding=1, bias=False)
class BasicBlock(nn.Module):
expansion = 1
def __init__(self, inplanes, planes, stride=1, downsample=None):
super(BasicBlock, self).__init__()
self.conv1 = conv3x3(inplanes, planes, stride)
#pooling 2
self.maxpool = nn.MaxPool2d(kernel_size=2, stride=2, padding=1)
self.relu = nn.ReLU(inplace=True)
self.conv2 = conv3x3(planes, planes)
self.maxpool = nn.MaxPool2d(kernel_size=2, stride=2, padding=1)
self.downsample = downsample
self.stride = stride
def forward(self, x):
residual = x
out = self.conv1(x)
out = self.maxpool(out)
out = self.relu(out)
out = self.conv2(out)
out = self.maxpool(out)
if self.downsample is not None:
residual = self.downsample(x)
out += residual
out = self.relu(out)
return out
class Convnet_Resnet_Layer(nn.Module):
def __init__(self, block, layers, num_classes=2):
self.inplanes = 1
super(Convnet_Resnet_Layer, self).__init__()
self.conv1 = nn.Conv2d(1, 16, kernel_size=3, stride=2, padding=3,
bias=False)
self.relu = nn.ReLU(inplace=True)
self.maxpool = nn.MaxPool2d(kernel_size=3, stride=2, padding=1)
self.layer1 = self._make_layer(block, 16, layers[0])
self.conv2 = nn.Conv2d(32, 32, kernel_size=3, stride=2, padding=3,
bias=False)
self.relu = nn.ReLU(inplace=True)
self.maxpool = nn.MaxPool2d(kernel_size=3, stride=2, padding=1)
#connect 64 on the last layer, with the 2 blocks
self.fc = nn.Linear(64 * block.expansion, num_classes)
# make sure that dimensions check out.
def _make_layer(self, block, planes, blocks, stride=1):
downsample = None
if stride != 1 or self.inplanes != planes * block.expansion:
downsample = nn.Sequential(
nn.Conv2d(self.inplanes, planes * block.expansion,
kernel_size=1, stride=stride, bias=False),
)
layers = []
layers.append(block(self.inplanes, planes, stride, downsample))
self.inplanes = planes * block.expansion
for i in range(1, blocks):
layers.append(block(self.inplanes, planes))
return nn.Sequential(*layers)
def forward(self, x):
x = self.conv1(x) # 224x224
x = self.relu(x)
x = self.maxpool(x) # 112x112
x = self.layer1(x) # 56x56
x = self.conv2(x)
x = self.relu(x)
x = self.maxpool(x)
x = self.layer2(x) # 28x28
# I did average pooling here. Can just omit if you want, but parameters will change
x = self.avgpool(x) # 1x1
x = x.view(x.size(0), -1)
x = self.out(x)
return x
# call the model and pass the list with the information of the blocks for each layer
model1 = Convnet_Resnet_Layer(BasicBlock, [2, 3,4 ])
print(model1)
结果如下:
Convnet_Resnet_Layer( (conv1):Conv2d(1,16,kernel_size =(3,3),步幅=(2,2),padding =(3,3),bias = False) (relu):ReLU(inplace) (maxpool):MaxPool2d(kernel_size = 3,stride = 2,padding = 1,dilation = 1,ceil_mode = False) (第1层):顺序( (0):基本块( (conv1):Conv2d(1,16,kernel_size =(3,3),步幅=(1,1),padding =(1,1),bias = False) (maxpool):MaxPool2d(kernel_size = 2,stride = 2,padding = 1,dilation = 1,ceil_mode = False) (relu):ReLU(inplace) (conv2):Conv2d(16,16,kernel_size =(3,3),步幅=(1,1),填充=(1,1),bias = False) (降采样):顺序( (0):Conv2d(1、16,kernel_size =(1、1),步幅=(1、1),bias = False) ) ) (1):BasicBlock( (conv1):Conv2d(16,16,kernel_size =(3,3),步幅=(1,1),填充=(1,1),bias = False) (maxpool):MaxPool2d(kernel_size = 2,stride = 2,padding = 1,dilation = 1,ceil_mode = False) (relu):ReLU(inplace) (conv2):Conv2d(16,16,kernel_size =(3,3),步幅=(1,1),填充=(1,1),bias = False) ) ) (conv2):Conv2d(32,32,kernel_size =(3,3),步幅=(2,2),padding =(3,3),bias = False) [fc):线性(in_features = 64,out_features = 2,bias = True) )
在(conv2)之后,我没有最大池或没有显示在模型中。
我将不胜感激。