首先看张核心的resnet层次结构图(图1),它诠释了resnet18-152是如何搭建的,其中resnet18和resnet34结构类似,而resnet50-resnet152结构类似。下面先看resnet18的源码
resnet18
首先是models.resnet18函数的调用
def resnet18(pretrained=False, **kwargs):
"""Constructs a ResNet-18 model.
Args:
pretrained (bool): If True, returns a model pre-trained on ImageNet
"""
#[2, 2, 2, 2]和结构图[]X2是对应的
model = ResNet(BasicBlock, [2, 2, 2, 2], **kwargs)
if pretrained: #加载模型权重
model.load_state_dict(model_zoo.load_url(model_urls['resnet18']))
return model
-
这里涉及到了一个BasicBlock类(resnet18和34),这样的一个结构我们称为一个block,因为在block内部的conv都使用了padding,输入的in_img_size和out_img_size都是56x56,在图2右边的shortcut只需要改变输入的channel的大小,输入bloack的输入tensor和输出tensor就可以相加(详细内容)
图2事实上图2是Bottleneck类(用于resnet50-152,稍后分析),其和BasicBlock差不多,图3为图2的精简版(ps:可以把下图视为为一个box_block,即多个block叠加在一起,x3说明有3个上图一样的结构串起来):
图3BasicBlock类,可以对比结构图中的resnet18和resnet34,类中expansion =1,其表示box_block中最后一个block的channel比上第一个block的channel,即:
def conv3x3(in_planes, out_planes, stride=1): "3x3 convolution with padding" return nn.Conv2d(in_planes, out_planes, kernel_size=3, stride=stride, padding=1, bias=False) class BasicBlock(nn.Module): expansion = 1 #inplanes其实就是channel,叫法不同 def __init__(self, inplanes, planes, stride=1, downsample=None): super(BasicBlock, self).__init__() self.conv1 = conv3x3(inplanes, planes, stride) self.bn1 = nn.BatchNorm2d(planes) self.relu = nn.ReLU(inplace=True) self.conv2 = conv3x3(planes, planes) self.bn2 = nn.BatchNorm2d(planes) self.downsample = downsample self.stride = stride 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) #把shortcut那的channel的维度统一 if self.downsample is not None: residual = self.downsample(x) out += residual out = self.relu(out)
接下来是ResNet类,其和我们通常定义的模型差不多一个init()+forward(),代码有点长,我们一步步来分析:
- 参考前面的结构图,所有的resnet的第一个conv层都是一样的,输出channel=64
- 然后到了self.layer1 = self._make_layer(block, 64, layers[0]),这里的layers[0]=2,然后我们进入到_make_layer函数,由于stride=1或当前的输入channel和上一个块的输出channel一样,因而可以直接相加
- self.layer2 = self._make_layer(block, 128, layers[1], stride=2),此时planes=128而self.inplanes=64为上box_block的输出channel,此时channel不一致,需要对输出的x扩维后才能相加,downsample 实现的就是该功能(ps:这里只有box_block中的第一个block需要downsample,为何?请看下图)
- self.layer3 = self._make_layer(block, 256, layers[2], stride=2),此时planes=256而self.inplanes=128为,此时也需要扩维后才能相加,layer4 同理。
图4
图4中下标2,3,4和上面的步骤对应,图中箭头旁数值表示box_block输入或者输出的channel数。
具体看图4-2,上一个box_block的最后一个block输出channel为64(也是下一个box_block的输入channel),而当前的box_block的第一个block的输出为128,在此需要扩维才能相加。然后到了当前box_block的第2个block,其输入channel和输出channel是一致的,因此无需扩维。
也就是说在box_block内部,只需要对第1个block进行扩维,因为在box_block内,第一个block输出channel和剩下的保持一致了。class ResNet(nn.Module): def __init__(self, block, layers, num_classes=1000): self.inplanes = 64 super(ResNet, self).__init__() self.conv1 = nn.Conv2d(3, 64, kernel_size=7, stride=2, padding=3, bias=False) self.bn1 = nn.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]) self.layer2 = self._make_layer(block, 128, layers[1], stride=2) self.layer3 = self._make_layer(block, 256, layers[2], stride=2) self.layer4 = self._make_layer(block, 512, layers[3], stride=2) self.avgpool = nn.AvgPool2d(7, stride=1) self.fc = nn.Linear(512 * block.expansion, num_classes) for m in self.modules(): if isinstance(m, nn.Conv2d): n = m.kernel_size[0] * m.kernel_size[1] * m.out_channels m.weight.data.normal_(0, math.sqrt(2. / n)) elif isinstance(m, nn.BatchNorm2d): m.weight.data.fill_(1) m.bias.data.zero_() def _make_layer(self, block, planes, blocks, stride=1): #downsample 主要用来处理H(x)=F(x)+x中F(x)和xchannel维度不匹配问题 downsample = None #self.inplanes为上个box_block的输出channel,planes为当前box_block块的输入channel 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), nn.BatchNorm2d(planes * block.expansion), ) 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) 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) x = self.avgpool(x) x = x.view(x.size(0), -1) x = self.fc(x)
resnet152
resnet152和resnet18差不多,Bottleneck类替换了BasicBlock,[3, 8, 36, 3]也和上面结构图对应。def resnet152(pretrained=False, **kwargs): """Constructs a ResNet-152 model. Args: pretrained (bool): If True, returns a model pre-trained on ImageNet """ model = ResNet(Bottleneck, [3, 8, 36, 3], **kwargs) if pretrained: model.load_state_dict(model_zoo.load_url(model_urls['resnet152']))
Bottleneck类,这里需要注意的是 expansion = 4,前面2个block的channel没有变,最后一个变成了第一个的4倍,具体可看本文的第2个图。
class Bottleneck(nn.Module): expansion = 4 def __init__(self, inplanes, planes, stride=1, downsample=None): super(Bottleneck, self).__init__() self.conv1 = nn.Conv2d(inplanes, planes, kernel_size=1, bias=False) self.bn1 = nn.BatchNorm2d(planes) self.conv2 = nn.Conv2d(planes, planes, kernel_size=3, stride=stride, padding=1, bias=False) self.bn2 = nn.BatchNorm2d(planes) self.conv3 = nn.Conv2d(planes, planes * 4, kernel_size=1, bias=False) self.bn3 = nn.BatchNorm2d(planes * 4) self.relu = nn.ReLU(inplace=True) self.downsample = downsample self.stride = stride 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) if self.downsample is not None: residual = self.downsample(x) out += residual out = self.relu(out)
图像输入大小问题:
首先pytorch输入的大小固定为224*224,超过这个大小就会报错,比如输入大小256*256RuntimeError: size mismatch, m1: [1 x 8192], m2: [2048 x 1000] at c:\miniconda2\conda-bld\pytorch-cpu_1519449358620\work\torch\lib\th\generic/THTensorMath.c:1434
- 1
- 1
首先我们看下,resnet在哪些地方改变了输出图像的大小
conv和pool层的输出大小都可以根据下面公式计算得出
但是resnet里面的卷积层太多了,就resnet152的height而言,其最后avgpool后的大小为,因此修改源码把图像的height和width传递进去,从而兼容非224的图片大小:
self.avgpool = nn.AvgPool2d(7, stride=1)
f = lambda x:math.ceil(x /32 - 7 + 1)
self.fc = nn.Linear(512 * block.expansion * f(w) * f(h), num_classes) #block.expansion=4
-
也可以在外面替换跳最后一个fc层,这里的2048即本文图1中resnet152对应的最后layer的输出channel,若是resnet18或resnet34则为512
model_ft = models.resnet152(pretrained=True) f = lambda x:math.ceil(x /32 - 7 + 1) model_ft.fc = nn.Linear(f(target_w) * f(target_h) * 2048, nb_classes)
-
还有另外一种暴力的方法,就是不管卷积层的输出大小,取其平均值做为输出,比如:
self.main = torchvision.models.resnet152(pretrained) self.main.avgpool = nn.AdaptiveAvgPool2d((1, 1))
-
第一次研究pytorch,请大神门轻喷
reference:
resnet详细介绍
deeper bottleneck Architectures详细介绍
-
-