Centernet network structure construction

One 、ResNet50 For the backbone network

1、 Backbone network part

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, stride=stride, bias=False) # change
        self.bn1 = nn.BatchNorm2d(planes)
        self.conv2 = nn.Conv2d(planes, planes, kernel_size=3, stride=1, # change
                    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: # Judge downsample Is it None
            residual = self.downsample(x)

        out += residual
        out = self.relu(out)

        return out

#-----------------------------------------------------------------#
#    Use Renset50 As a backbone feature extraction network , Input is 512x512x3 You will eventually get one 
#   16x16x2048 Effective feature layer 
#-----------------------------------------------------------------#
class ResNet(nn.Module):
    def __init__(self, block, layers, num_classes=1000):
        self.inplanes = 64
        super(ResNet, self).__init__()
        # 512,512,3 -> 256,256,64
        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)
        # 256x256x64 -> 128x128x64
        self.maxpool = nn.MaxPool2d(kernel_size=3, stride=2, padding=0, ceil_mode=True) # change

        # 128x128x64 -> 128x128x256
        self.layer1 = self._make_layer(block, 64, layers[0])

        # 128x128x256 -> 64x64x512
        self.layer2 = self._make_layer(block, 128, layers[1], stride=2)

        # 64x64x512 -> 32x32x1024
        self.layer3 = self._make_layer(block, 256, layers[2], stride=2)  

        # 32x32x1024 -> 16x16x2048
        self.layer4 = self._make_layer(block, 512, layers[3], stride=2)
        
        self.avgpool = nn.AvgPool2d(7)
        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 = None
        if stride != 1 or self.inplanes != planes * block.expansion:
            downsample = nn.Sequential(     # structure downsample, Contains a Conv2d And a BatchNorm2d
                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)) #downsample Pass in block As the residual edge 
        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)

        return x

def resnet50(pretrained = True):
    model = ResNet(Bottleneck, [3, 4, 6, 3])
    if pretrained:
        state_dict = load_state_dict_from_url(model_urls['resnet50'], model_dir = 'model_data/')
        model.load_state_dict(state_dict)
    #----------------------------------------------------------#
    #    Get the feature extraction part 
    #    take layer4 Layer and before 
    features = list([model.conv1, model.bn1, model.relu, model.maxpool, model.layer1, model.layer2, model.layer3, model.layer4])
    features = nn.Sequential(*features)
    return features

 2、 The backbone network is followed by three deconvolutions for up sampling

#  Three after building the backbone network ConvTranspose2d Deconvolution for up sampling 
class resnet50_Decoder(nn.Module):
    def __init__(self, inplanes, bn_momentum=0.1):
        super(resnet50_Decoder, self).__init__()
        self.bn_momentum = bn_momentum
        self.inplanes = inplanes
        self.deconv_with_bias = False
        
        #----------------------------------------------------------#
        #   16,16,2048 -> 32,32,256 -> 64,64,128 -> 128,128,64
        #    utilize ConvTranspose2d Perform three times of deconvolution up sampling .
        #    Each time the width and height of the feature layer become twice the original .
        #----------------------------------------------------------#
        self.deconv_layers = self._make_deconv_layer(
            num_layers=3,
            num_filters=[256, 128, 64],
            num_kernels=[4, 4, 4],
        )

    def _make_deconv_layer(self, num_layers, num_filters, num_kernels):
        layers = []
        for i in range(num_layers):
            kernel = num_kernels[i]
            planes = num_filters[i]

            layers.append(
                nn.ConvTranspose2d(
                    in_channels=self.inplanes, # 2048 256 128
                    out_channels=planes, # 256 128 64
                    kernel_size=kernel, # 4 4 4
                    stride=2,
                    padding=1,
                    output_padding=0,
                    bias=self.deconv_with_bias))
            layers.append(nn.BatchNorm2d(planes, momentum=self.bn_momentum))
            layers.append(nn.ReLU(inplace=True))
            self.inplanes = planes
        return nn.Sequential(*layers)

    def forward(self, x): # Positive propagation process 
        return self.deconv_layers(x)

 3、 Detection head part  

class resnet50_Head(nn.Module):
    def __init__(self, num_classes=80, channel=64, bn_momentum=0.1):
        super(resnet50_Head, self).__init__()
        #-----------------------------------------------------------------#
        #    Upsampling the acquired features , Make classification prediction and regression prediction 
        #   128, 128, 64 -> 128, 128, 64 -> 128, 128, num_classes
        #                -> 128, 128, 64 -> 128, 128, 2
        #                -> 128, 128, 64 -> 128, 128, 2
        #-----------------------------------------------------------------#
        #  Thermodynamic diagram prediction part 
        #  The result of thermodynamic map prediction can represent the result of network classification prediction 
        self.cls_head = nn.Sequential(
            nn.Conv2d(64, channel,
                      kernel_size=3, stride=1, padding=1, bias=False),
            nn.BatchNorm2d(channel, momentum=bn_momentum),
            nn.ReLU(inplace=True), #ConvBNReLU It can be regarded as a function of feature integration 
            nn.Conv2d(channel, num_classes,
                      kernel_size=1, stride=1, padding=0)) #1*1 Convolution to adjust the number of channels 
        #  Part of width and height prediction 
        #  Number of output channels 2 It is equivalent to the width and height of the object at each feature point 
        self.wh_head = nn.Sequential(
            nn.Conv2d(64, channel,
                      kernel_size=3, stride=1, padding=1, bias=False),
            nn.BatchNorm2d(channel, momentum=bn_momentum),
            nn.ReLU(inplace=True), #ConvBNReLU It can be regarded as a function of feature integration 
            nn.Conv2d(channel, 2,
                      kernel_size=1, stride=1, padding=0)) #1*1 Convolution to adjust the number of channels 

        #  Part of the central point prediction 
        #  Number of output channels 2 Equivalent to the center point at X、Y The offset of the axis 
        self.reg_head = nn.Sequential(
            nn.Conv2d(64, channel,
                      kernel_size=3, stride=1, padding=1, bias=False),
            nn.BatchNorm2d(channel, momentum=bn_momentum),
            nn.ReLU(inplace=True), #ConvBNReLU It can be regarded as a function of feature integration 
            nn.Conv2d(channel, 2,
                      kernel_size=1, stride=1, padding=0)) #1*1 Convolution to adjust the number of channels 

    def forward(self, x):
        hm = self.cls_head(x).sigmoid_()
        wh = self.wh_head(x)
        offset = self.reg_head(x)
        return hm, wh, offset

4、CenterNet_Resnet50 The overall structure

class CenterNet_Resnet50(nn.Module):
    def __init__(self, num_classes = 20, pretrained = False):
        super(CenterNet_Resnet50, self).__init__()
        # 512,512,3 -> 16,16,2048
        self.backbone = resnet50(pretrained = pretrained)
        # 16,16,2048 -> 128,128,64
        self.decoder = resnet50_Decoder(2048)
        #-----------------------------------------------------------------#
        #    Upsampling the acquired features , Make classification prediction and regression prediction 
        #   128, 128, 64 -> 128, 128, 64 -> 128, 128, num_classes
        #                -> 128, 128, 64 -> 128, 128, 2
        #                -> 128, 128, 64 -> 128, 128, 2
        #-----------------------------------------------------------------#
        self.head = resnet50_Head(channel=64, num_classes=num_classes)

    
    def forward(self, x):
        feat = self.backbone(x)
        return self.head(self.decoder(feat))

reference

Pytorch Build your own Centernet Target detection platform （Bubbliiiing Deep learning course ）_ Bili, Bili _bilibili