Pytoch official fast r-cnn source code analysis (I) -- feature extraction

In depth Pytorch official Faster R-CNN Source code , Bloggers will explain every code in as much detail as possible , If it helps you, you can pay attention to it and praise it , There are questions in the comments section , Bloggers will try their best to answer .

---

Faster R-CNN Thesis link
Pytorch official Faster R-CNN Link to the code documentation for .

Pytorch The official example code is as follows ：

import torch
import torchvision

model = torchvision.models.detection.fasterrcnn_resnet50_fpn(pretrained=True)
# For training
images, boxes = torch.rand(4, 3, 600, 1200), torch.rand(4, 11, 4)
boxes[:, :, 2:4] = boxes[:, :, 0:2] + boxes[:, :, 2:4]
labels = torch.randint(1, 91, (4, 11))
images = list(image for image in images)
targets = []
for i in range(len(images)):
    d = {
    'boxes': boxes[i], 'labels': labels[i]}
    targets.append(d)
output = model(images, targets)
# For inference
model.eval()
x = [torch.rand(3, 300, 400), torch.rand(3, 500, 400)]
predictions = model(x)

# optionally, if you want to export the model to ONNX:
torch.onnx.export(model, x, "faster_rcnn.onnx", opset_version = 11)

The following is a detailed description of the sample code .

---

First , initialization Faster R-CNN Model .

model = torchvision.models.detection.fasterrcnn_resnet50_fpn(pretrained=True)

It can be seen that , The backbone network is used here Resnet-50-FPN Of Faster R-CNN. Next Debug Enter the internal code .

def fasterrcnn_resnet50_fpn(pretrained=False, progress=True,
                            num_classes=91, pretrained_backbone=True, trainable_backbone_layers=3, **kwargs):
    """ Constructs a Faster R-CNN model with a ResNet-50-FPN backbone.  Building a backbone network is  ResNet-50-FPN  Of  Faster R-CNN  Model . The input to the model is expected to be a list of tensors, each of shape ``[C, H, W]``, one for each image, and should be in ``0-1`` range. Different images can have different sizes.  The input to the model should be an input from tensors A list of components , Every tensor The shape of is [C,H,W], For each image, the element value should be in [0,1] Within the scope of , Different images have different sizes . The behavior of the model changes depending if it is in training or evaluation mode.  The model has two modes: training and evaluation , The performance of the model depends on the model . During training, the model expects both the input tensors, as well as a targets (list of dictionary), containing: - boxes (``FloatTensor[N, 4]``): the ground-truth boxes in ``[x1, y1, x2, y2]`` format, with values of ``x`` between ``0`` and ``W`` and values of ``y`` between ``0`` and ``H`` - labels (``Int64Tensor[N]``): the class label for each ground-truth box  In the process of training , The model needs to input the tensor, And the goal ( A list of dictionaries ), It contains ： -  Frame （FloatTensor[N,4]）： The real box is [x1,y1,x2,y2] In the form of ,x  The value of the  0~W  Between ,y  The value of the  0-H  Between . -  label （Int64Tensor[N]）： Category labels for each real box . The model returns a ``Dict[Tensor]`` during training, containing the classification and regression losses for both the RPN and the R-CNN.  During training , The model returns a  ”Dict[Tensor]“, contain  RPN  And  R-CNN  Classification of stages and regression losses . During inference, the model requires only the input tensors, and returns the post-processed predictions as a ``List[Dict[Tensor]]``, one for each input image. The fields of the ``Dict`` are as follows: - boxes (``FloatTensor[N, 4]``): the predicted boxes in ``[x1, y1, x2, y2]`` format, with values of ``x`` between ``0`` and ``W`` and values of ``y`` between ``0`` and ``H`` - labels (``Int64Tensor[N]``): the predicted labels for each image - scores (``Tensor[N]``): the scores or each prediction  In the process of reasoning , The model only needs to input the tensor, Then return the post-processing prediction results to  "List[Dict[Tensor]]"  In the form of , For each input image , Its  "Dict"  The fields are as follows ： -  Frame （FloatTensor[N,4]）： The prediction box is [x1,y1,x2,y2] In the form of ,x  The value of the  0~W  Between ,y  The value of the  0~H  Between . -  label （Int64Tensor[N]）： Prediction labels for each image . -  fraction （Tensor[N]）： Score per forecast . Faster R-CNN is exportable to ONNX for a fixed batch size with inputs images of fixed size. Faster R—CNN  Can be exported as a fixed batch size field fixed size input image  ONNX  Format . Arguments: pretrained (bool): If True, returns a model pre-trained on COCO train2017 progress (bool): If True, displays a progress bar of the download to stderr pretrained_backbone (bool): If True, returns a model with backbone pre-trained on Imagenet num_classes (int): number of output classes of the model (including the background) trainable_backbone_layers (int): number of trainable (not frozen) resnet layers starting from final block. Valid values are between 0 and 5, with 5 meaning all backbone layers are trainable.  Parameters ： pretrianed（bool）： If it is true , Return to a  COCO train2017  The pre training model on . progress（bool）： If it is true , Display the download progress bar on the screen . pretrained_backbone（bool）： If it is true , Return to a  Imagenet  Backbone network pre training model on . num_classes（int）： Number of types of model output （ Including background ）. trainable_backbone_layers（int）： Start with the last block  ResNet  Number of layers （ Not frozen ）. The legal value is  0~5  Between ,5  This means that all layers of the backbone network are trainable . """
	#  Use  assert  Judge  trainable_backbone_layers  Whether the value of is legal 
    assert trainable_backbone_layers <= 5 and trainable_backbone_layers >= 0 
    # dont freeze any layers if pretrained model or backbone is not used
    #  If the pre training model or the pre training backbone network is not used , Do not freeze any layers .
    if not (pretrained or pretrained_backbone):
        trainable_backbone_layers = 5
    if pretrained:
        # no need to download the backbone if pretrained is set
        #  If the pre training model is used , There is no need to download the pre training backbone 
        pretrained_backbone = False
   	#  obtain  ResNet_FPN  Backbone network 
    backbone = resnet_fpn_backbone('resnet50', pretrained_backbone, trainable_layers=trainable_backbone_layers)
    #  obtain  Faster R-CNN  Model 
    model = FasterRCNN(backbone, num_classes, **kwargs)
    if pretrained:
        #  If you use a pre training model , Download the relevant pre training model configuration 
        state_dict = load_state_dict_from_url(model_urls['fasterrcnn_resnet50_fpn_coco'],
                                              progress=progress)
        #  Load the model configuration into the model 
        model.load_state_dict(state_dict)
    return model #  Back to the model 

Debug Access ResNet_FPN Corresponding code of backbone network .

def resnet_fpn_backbone(
    backbone_name,
    pretrained,
    norm_layer=misc_nn_ops.FrozenBatchNorm2d,
    trainable_layers=3,
    returned_layers=None,
    extra_blocks=None
):
    """ Constructs a specified ResNet backbone with FPN on top. Freezes the specified number of layers in the backbone.  Build one that adds... At the top FPN Of ResNet Backbone network . Freeze the specified number of layers in the backbone network . Examples:: >>> from torchvision.models.detection.backbone_utils import resnet_fpn_backbone >>> backbone = resnet_fpn_backbone('resnet50', pretrained=True, trainable_layers=3) >>> # get some dummy image >>> x = torch.rand(1,3,64,64) >>> # compute the output >>> output = backbone(x) >>> print([(k, v.shape) for k, v in output.items()]) >>> # returns >>> [('0', torch.Size([1, 256, 16, 16])), >>> ('1', torch.Size([1, 256, 8, 8])), >>> ('2', torch.Size([1, 256, 4, 4])), >>> ('3', torch.Size([1, 256, 2, 2])), >>> ('pool', torch.Size([1, 256, 1, 1]))] Arguments: backbone_name (string): resnet architecture. Possible values are 'ResNet', 'resnet18', 'resnet34', 'resnet50', 'resnet101', 'resnet152', 'resnext50_32x4d', 'resnext101_32x8d', 'wide_resnet50_2', 'wide_resnet101_2' norm_layer (torchvision.ops): it is recommended to use the default value. For details visit: (https://github.com/facebookresearch/maskrcnn-benchmark/issues/267) pretrained (bool): If True, returns a model with backbone pre-trained on Imagenet trainable_layers (int): number of trainable (not frozen) resnet layers starting from final block. Valid values are between 0 and 5, with 5 meaning all backbone layers are trainable.  Parameters ： backbone_name (string)：resnet  framework . The possible value is  'ResNet','resnet18','resnet34','resnet50','resnet101','resnet152', 'resnext50_32x4d','resnet101_32x8d','wide_resnet50_2','wide_resnet101_2' norm_layer (torchivision.ops)： It is recommended to use the default value . For details, please visit ： (https://github.com/facebookresearch/maskrcnn-benchmark/issues/267) pretrained (bool)： If it is true , Return to a  Imagenet  Pre training backbone network model on  trainable_layers (int)： Start with the last block  ResNet  Number of layers （ Not frozen ）. The legal value is  0~5  Between ,5  This means that all layers of the backbone network are trainable . """
    backbone = resnet.__dict__[backbone_name](
        pretrained=pretrained,
        norm_layer=norm_layer) #  obtain resnet-50 Backbone network 

    # select layers that wont be frozen
    #  Select the frozen layer （ Don't take part in training ）
    assert trainable_layers <= 5 and trainable_layers >= 0
    layers_to_train = ['layer4', 'layer3', 'layer2', 'layer1', 'conv1'][:trainable_layers]
    # freeze layers only if pretrained backbone is used
    #  The layer is frozen only when the pre training backbone network is used 
    for name, parameter in backbone.named_parameters():
        if all([not name.startswith(layer) for layer in layers_to_train]):
            parameter.requires_grad_(False)

    if extra_blocks is None:
        extra_blocks = LastLevelMaxPool()

    if returned_layers is None:
        returned_layers = [1, 2, 3, 4]
    assert min(returned_layers) > 0 and max(returned_layers) < 5
    return_layers = {
    f'layer{
      k}': str(v) for v, k in enumerate(returned_layers)}

    in_channels_stage2 = backbone.inplanes // 8
    in_channels_list = [in_channels_stage2 * 2 ** (i - 1) for i in returned_layers]
    out_channels = 256
    return BackboneWithFPN(backbone, return_layers, in_channels_list, out_channels, extra_blocks=extra_blocks)

Debug Access ResNet-50 The code of the backbone network .

class ResNet(nn.Module):

    def __init__(self, block, layers, num_classes=1000, zero_init_residual=False,
                 groups=1, width_per_group=64, replace_stride_with_dilation=None,
                 norm_layer=None):
        super(ResNet, self).__init__()
        if norm_layer is None:
            norm_layer = nn.BatchNorm2d
        self._norm_layer = norm_layer

        self.inplanes = 64
        self.dilation = 1
        if replace_stride_with_dilation is None:
            # each element in the tuple indicates if we should replace
            # the 2x2 stride with a dilated convolution instead
            replace_stride_with_dilation = [False, False, False]
        if len(replace_stride_with_dilation) != 3:
            raise ValueError("replace_stride_with_dilation should be None "
                             "or a 3-element tuple, got {}".format(replace_stride_with_dilation))
        self.groups = groups
        self.base_width = width_per_group
        self.conv1 = nn.Conv2d(3, self.inplanes, kernel_size=7, stride=2, padding=3,
                               bias=False)
        self.bn1 = norm_layer(self.inplanes)
        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,
                                       dilate=replace_stride_with_dilation[0])
        self.layer3 = self._make_layer(block, 256, layers[2], stride=2,
                                       dilate=replace_stride_with_dilation[1])
        self.layer4 = self._make_layer(block, 512, layers[3], stride=2,
                                       dilate=replace_stride_with_dilation[2])
        self.avgpool = nn.AdaptiveAvgPool2d((1, 1))
        self.fc = nn.Linear(512 * block.expansion, num_classes)
        
    def _make_layer(self, block, planes, blocks, stride=1, dilate=False):
        norm_layer = self._norm_layer
        downsample = None
        previous_dilation = self.dilation
        if dilate:
            self.dilation *= stride
            stride = 1
        if stride != 1 or self.inplanes != planes * block.expansion:
            downsample = nn.Sequential(
                conv1x1(self.inplanes, planes * block.expansion, stride),
                norm_layer(planes * block.expansion),
            )

        layers = []
        layers.append(block(self.inplanes, planes, stride, downsample, self.groups,
                            self.base_width, previous_dilation, norm_layer))
        self.inplanes = planes * block.expansion
        for _ in range(1, blocks):
            layers.append(block(self.inplanes, planes, groups=self.groups,
                                base_width=self.base_width, dilation=self.dilation,
                                norm_layer=norm_layer))

        return nn.Sequential(*layers)

    def _forward_impl(self, x):
        # See note [TorchScript super()]
        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 = torch.flatten(x, 1)
        x = self.fc(x)

        return x

    def forward(self, x):
        return self._forward_impl(x)

ResNet-50 The network structure of is shown in the figure below .

ResNet-50 Adopted BottleNeck structure , Their ratio BasicNeck Save more parameters . The code is as follows ：

class Bottleneck(nn.Module):
    # Bottleneck in torchvision places the stride for downsampling at 3x3 convolution(self.conv2)
    # while original implementation places the stride at the first 1x1 convolution(self.conv1)
    # according to "Deep residual learning for image recognition"https://arxiv.org/abs/1512.03385.
    # This variant is also known as ResNet V1.5 and improves accuracy according to
    # https://ngc.nvidia.com/catalog/model-scripts/nvidia:resnet_50_v1_5_for_pytorch.

    expansion = 4

    def __init__(self, inplanes, planes, stride=1, downsample=None, groups=1,
                 base_width=64, dilation=1, norm_layer=None):
        super(Bottleneck, self).__init__()
        if norm_layer is None:
            norm_layer = nn.BatchNorm2d
        width = int(planes * (base_width / 64.)) * groups
        # Both self.conv2 and self.downsample layers downsample the input when stride != 1
        self.conv1 = conv1x1(inplanes, width)
        self.bn1 = norm_layer(width)
        self.conv2 = conv3x3(width, width, stride, groups, dilation)
        self.bn2 = norm_layer(width)
        self.conv3 = conv1x1(width, planes * self.expansion)
        self.bn3 = norm_layer(planes * self.expansion)
        self.relu = nn.ReLU(inplace=True)
        self.downsample = downsample
        self.stride = stride

    def forward(self, x):
        identity = 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:
            identity = self.downsample(x)

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

        return out

At last we got ResNet-50 Network structure ：

ResNet(
  (conv1): Conv2d(3, 64, kernel_size=(7, 7), stride=(2, 2), padding=(3, 3), bias=False)
  (bn1): FrozenBatchNorm2d(64)
  (relu): ReLU(inplace=True)
  (maxpool): MaxPool2d(kernel_size=3, stride=2, padding=1, dilation=1, ceil_mode=False)
  (layer1): Sequential(
    (0): Bottleneck(
      (conv1): Conv2d(64, 64, kernel_size=(1, 1), stride=(1, 1), bias=False)
      (bn1): FrozenBatchNorm2d(64)
      (conv2): Conv2d(64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)
      (bn2): FrozenBatchNorm2d(64)
      (conv3): Conv2d(64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False)
      (bn3): FrozenBatchNorm2d(256)
      (relu): ReLU(inplace=True)
      (downsample): Sequential(
        (0): Conv2d(64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False)
        (1): FrozenBatchNorm2d(256)
      )
    )
    (1): Bottleneck(
      (conv1): Conv2d(256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False)
      (bn1): FrozenBatchNorm2d(64)
      (conv2): Conv2d(64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)
      (bn2): FrozenBatchNorm2d(64)
      (conv3): Conv2d(64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False)
      (bn3): FrozenBatchNorm2d(256)
      (relu): ReLU(inplace=True)
    )
    (2): Bottleneck(
      (conv1): Conv2d(256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False)
      (bn1): FrozenBatchNorm2d(64)
      (conv2): Conv2d(64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)
      (bn2): FrozenBatchNorm2d(64)
      (conv3): Conv2d(64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False)
      (bn3): FrozenBatchNorm2d(256)
      (relu): ReLU(inplace=True)
    )
  )
  (layer2): Sequential(
    (0): Bottleneck(
      (conv1): Conv2d(256, 128, kernel_size=(1, 1), stride=(1, 1), bias=False)
      (bn1): FrozenBatchNorm2d(128)
      (conv2): Conv2d(128, 128, kernel_size=(3, 3), stride=(2, 2), padding=(1, 1), bias=False)
      (bn2): FrozenBatchNorm2d(128)
      (conv3): Conv2d(128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False)
      (bn3): FrozenBatchNorm2d(512)
      (relu): ReLU(inplace=True)
      (downsample): Sequential(
        (0): Conv2d(256, 512, kernel_size=(1, 1), stride=(2, 2), bias=False)
        (1): FrozenBatchNorm2d(512)
      )
    )
    (1): Bottleneck(
      (conv1): Conv2d(512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False)
      (bn1): FrozenBatchNorm2d(128)
      (conv2): Conv2d(128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)
      (bn2): FrozenBatchNorm2d(128)
      (conv3): Conv2d(128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False)
      (bn3): FrozenBatchNorm2d(512)
      (relu): ReLU(inplace=True)
    )
    (2): Bottleneck(
      (conv1): Conv2d(512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False)
      (bn1): FrozenBatchNorm2d(128)
      (conv2): Conv2d(128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)
      (bn2): FrozenBatchNorm2d(128)
      (conv3): Conv2d(128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False)
      (bn3): FrozenBatchNorm2d(512)
      (relu): ReLU(inplace=True)
    )
    (3): Bottleneck(
      (conv1): Conv2d(512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False)
      (bn1): FrozenBatchNorm2d(128)
      (conv2): Conv2d(128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)
      (bn2): FrozenBatchNorm2d(128)
      (conv3): Conv2d(128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False)
      (bn3): FrozenBatchNorm2d(512)
      (relu): ReLU(inplace=True)
    )
  )
  (layer3): Sequential(
    (0): Bottleneck(
      (conv1): Conv2d(512, 256, kernel_size=(1, 1), stride=(1, 1), bias=False)
      (bn1): FrozenBatchNorm2d(256)
      (conv2): Conv2d(256, 256, kernel_size=(3, 3), stride=(2, 2), padding=(1, 1), bias=False)
      (bn2): FrozenBatchNorm2d(256)
      (conv3): Conv2d(256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False)
      (bn3): FrozenBatchNorm2d(1024)
      (relu): ReLU(inplace=True)
      (downsample): Sequential(
        (0): Conv2d(512, 1024, kernel_size=(1, 1), stride=(2, 2), bias=False)
        (1): FrozenBatchNorm2d(1024)
      )
    )
    (1): Bottleneck(
      (conv1): Conv2d(1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False)
      (bn1): FrozenBatchNorm2d(256)
      (conv2): Conv2d(256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)
      (bn2): FrozenBatchNorm2d(256)
      (conv3): Conv2d(256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False)
      (bn3): FrozenBatchNorm2d(1024)
      (relu): ReLU(inplace=True)
    )
    (2): Bottleneck(
      (conv1): Conv2d(1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False)
      (bn1): FrozenBatchNorm2d(256)
      (conv2): Conv2d(256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)
      (bn2): FrozenBatchNorm2d(256)
      (conv3): Conv2d(256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False)
      (bn3): FrozenBatchNorm2d(1024)
      (relu): ReLU(inplace=True)
    )
    (3): Bottleneck(
      (conv1): Conv2d(1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False)
      (bn1): FrozenBatchNorm2d(256)
      (conv2): Conv2d(256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)
      (bn2): FrozenBatchNorm2d(256)
      (conv3): Conv2d(256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False)
      (bn3): FrozenBatchNorm2d(1024)
      (relu): ReLU(inplace=True)
    )
    (4): Bottleneck(
      (conv1): Conv2d(1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False)
      (bn1): FrozenBatchNorm2d(256)
      (conv2): Conv2d(256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)
      (bn2): FrozenBatchNorm2d(256)
      (conv3): Conv2d(256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False)
      (bn3): FrozenBatchNorm2d(1024)
      (relu): ReLU(inplace=True)
    )
    (5): Bottleneck(
      (conv1): Conv2d(1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False)
      (bn1): FrozenBatchNorm2d(256)
      (conv2): Conv2d(256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)
      (bn2): FrozenBatchNorm2d(256)
      (conv3): Conv2d(256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False)
      (bn3): FrozenBatchNorm2d(1024)
      (relu): ReLU(inplace=True)
    )
  )
  (layer4): Sequential(
    (0): Bottleneck(
      (conv1): Conv2d(1024, 512, kernel_size=(1, 1), stride=(1, 1), bias=False)
      (bn1): FrozenBatchNorm2d(512)
      (conv2): Conv2d(512, 512, kernel_size=(3, 3), stride=(2, 2), padding=(1, 1), bias=False)
      (bn2): FrozenBatchNorm2d(512)
      (conv3): Conv2d(512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False)
      (bn3): FrozenBatchNorm2d(2048)
      (relu): ReLU(inplace=True)
      (downsample): Sequential(
        (0): Conv2d(1024, 2048, kernel_size=(1, 1), stride=(2, 2), bias=False)
        (1): FrozenBatchNorm2d(2048)
      )
    )
    (1): Bottleneck(
      (conv1): Conv2d(2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False)
      (bn1): FrozenBatchNorm2d(512)
      (conv2): Conv2d(512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)
      (bn2): FrozenBatchNorm2d(512)
      (conv3): Conv2d(512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False)
      (bn3): FrozenBatchNorm2d(2048)
      (relu): ReLU(inplace=True)
    )
    (2): Bottleneck(
      (conv1): Conv2d(2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False)
      (bn1): FrozenBatchNorm2d(512)
      (conv2): Conv2d(512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)
      (bn2): FrozenBatchNorm2d(512)
      (conv3): Conv2d(512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False)
      (bn3): FrozenBatchNorm2d(2048)
      (relu): ReLU(inplace=True)
    )
  )
  (avgpool): AdaptiveAvgPool2d(output_size=(1, 1))
  (fc): Linear(in_features=2048, out_features=1000, bias=True)
)

After that, it was added as FPN.

As shown in the figure above ,FPN It is possible to build feature pyramids with strong semantics at various scales , The specific principle can be seen in This blog .FPN Get... Here ResNet-50 The feature map extracted at each stage plus the last feature map of the maximum pool is a total of five feature maps . The code is as follows ：

    def forward(self, x: Dict[str, Tensor]) -> Dict[str, Tensor]:
        """ Computes the FPN for a set of feature maps. Arguments: x (OrderedDict[Tensor]): feature maps for each feature level. Returns: results (OrderedDict[Tensor]): feature maps after FPN layers. They are ordered from highest resolution first. """
        # unpack OrderedDict into two lists for easier handling
        names = list(x.keys())
        x = list(x.values())

        last_inner = self.get_result_from_inner_blocks(x[-1], -1)
        results = []
        results.append(self.get_result_from_layer_blocks(last_inner, -1))

        for idx in range(len(x) - 2, -1, -1):
            inner_lateral = self.get_result_from_inner_blocks(x[idx], idx) # 1x1  Convolution reduces the number of channels to 256
            feat_shape = inner_lateral.shape[-2:]
            inner_top_down = F.interpolate(last_inner, size=feat_shape, mode="nearest") #  On the sampling 
            last_inner = inner_lateral + inner_top_down #  Transverse connection 
            results.insert(0, self.get_result_from_layer_blocks(last_inner, idx)) # 3x3  Convolution to eliminate aliasing effect 

        if self.extra_blocks is not None:
            results, names = self.extra_blocks(results, x, names) #  Maximize the pool to obtain the characteristic map of the fifth layer 

        # make it back an OrderedDict
        out = OrderedDict([(k, v) for k, v in zip(names, results)])

        return out

thus , Feature map extraction has been completed , Next, we will carry on RPN( Generation of regions of interest ).