class Bottleneck(nn.Module):
//一个残差块的结构。第一个1*1卷积会缩小通道数,第二个3*3卷积,第三个1*1卷积会变回原来的通道数
expansion = 4
//inplanes:输入通道; planes*expansion:输出通道;
def __init__(self, inplanes, planes, stride=1, dilation=1, downsample=None):
super(Bottleneck, self).__init__()
//第一个1*1卷积,stride一定为1
self.conv1 = nn.Conv2d(inplanes, planes, kernel_size=1, bias=False)
self.bn1 = BatchNorm2d(planes)
//第二个3*3卷积,stride不一定为1,可以有空洞
self.conv2 = nn.Conv2d(planes, planes, kernel_size=3, stride=stride,
dilation=dilation, padding=dilation, bias=False)
self.bn2 = BatchNorm2d(planes)
//第三个1*1卷积,stride一定为1,通道数扩大4倍
self.conv3 = nn.Conv2d(planes, planes * 4, kernel_size=1, bias=False)
self.bn3 = BatchNorm2d(planes * 4)
self.relu = nn.ReLU(inplace=True)
//下采样通道数
self.downsample = downsample
self.stride = stride
self.dilation = dilation
def forward(self, x):
residual = x
out = self.conv1(x)
out = self.bn1(out)
out = self.relu(out)
out = self.conv2(out)
out = self.bn2(out)
out = self.relu(out)
out = self.conv3(out)
out = self.bn3(out)
//如果输入residual与经过1*1,3*3,1*1之后的输出out通道数不同,对输入residual进行下采样,再与输出相加
if self.downsample is not None:
residual = self.downsample(x)
out += residual
out = self.relu(out)
return out
class ResNet(nn.Module):
//nInputChannels为输入通道,block为待传的Bottleneck
def __init__(self, nInputChannels, block, layers, os=16, pretrained=False):
//self.inplanes是传给self._make_layer函数的输入通道数
self.inplanes = 64
super(ResNet, self).__init__()
//在不同的os下,四个layer的参数,output stride为该矩阵经过多次卷积pooling操作后,尺寸缩小的值,例如:
input image为224 * 224,经过多次卷积pooling操作后,feature map为7*7,那么output stride为224/7 = 32.
//这里的blocks指的是每个layer中的残差块个数
if os == 16:
strides = [1, 2, 2, 1]
dilations = [1, 1, 1, 2]
blocks = [1, 2, 4]
elif os == 8:
strides = [1, 2, 1, 1]
dilations = [1, 1, 2, 2]
blocks = [1, 2, 1]
else:
raise NotImplementedError
# Modules
//输入通道数为nInputChannels,输出为64
self.conv1 = nn.Conv2d(nInputChannels, 64, kernel_size=7, stride=2, padding=3,
bias=False)
self.bn1 = BatchNorm2d(64)
self.relu = nn.ReLU(inplace=True)
self.maxpool = nn.MaxPool2d(kernel_size=3, stride=2, padding=1)
self.layer1 = self._make_layer(block, 64, layers[0], stride=strides[0], dilation=dilations[0])
self.layer2 = self._make_layer(block, 128, layers[1], stride=strides[1], dilation=dilations[1])
self.layer3 = self._make_layer(block, 256, layers[2], stride=strides[2], dilation=dilations[2])
self.layer4 = self._make_layer(block, 512, blocks=blocks, stride=strides[3], dilation=dilations[3])
self._init_weight()
if pretrained:
self._load_pretrained_model()
//block是待传递的bottlenet,planes为输入通道数,不同的layers分别为64、128、156,blocks指的是每个layer中的残差块
//个数,不为1的stride传入会用某一个layer每个bottlenet中的3*3卷积进行下采样
def _make_layer(self, block, planes, blocks, stride=1, dilation=1):
downsample = None
//如果马上使用的bottlenet中3*3卷积的stride不是1(意思是需要在这个layer的第一层进行一次图片尺寸的下采样)或者
//输出通道数planes*expansion与inplanes不同(意思是如果每个layer的输出通道数planes*expansion和输入通道数不同
//(输入通道数第一次是64,第二次是64*4,第三次是128*4)
if stride != 1 or self.inplanes != planes * block.expansion:
//这个函数在这里没有使用,只是定义了一下。对图片尺寸进行下采样,另外把输入通道数inplanes变成输出通道数
//planes * block.expansion
downsample = nn.Sequential(
nn.Conv2d(self.inplanes, planes * block.expansion,
kernel_size=1, stride=stride, bias=False),
BatchNorm2d(planes * block.expansion),
)
layers = []
//在这里调用第一个Bottleneck(),形成第一个block,每个layer是多个block组成,但是只对第一个block传入stride
//(只对第一个block可能进行下采样),当layer的第一个block的输入通道inplanes不等于输出通道planes*expansion,
//进行downsample。第一层输入为64,经过1*1卷积,3*3卷积,1*1卷积后输出64*4,输入和输出不同维度
//(另外如果输出是经过了stride下采样,输出图片尺寸也和输入不同)不能相加,所以在downsample中对输入进行
//stride下采样,对输入图片的通道数进行升维,从64变到64*4
layers.append(block(self.inplanes, planes, stride, dilation, downsample))
self.inplanes = planes * block.expansion
//对每个layer中的除过第一个bottlenet外的bottlenet们,stride为默认值1,不进行下采样,另外因为已经定义了
//self.inplanes = planes * block.expansion,所以剩余的这些block的输入通道和输出通道是一样的,所以不再使用
//downsample()
for i in range(1, blocks):
layers.append(block(self.inplanes, planes))
return nn.Sequential(*layers)
def forward(self, input):
x = self.conv1(input)
x = self.bn1(x)
x = self.relu(x)
x = self.maxpool(x)
x = self.layer1(x)
x = self.layer2(x)
x = self.layer3(x)
x = self.layer4(x)
return x