Pytorch官方Faster R-CNN源代码解析(一)——特征提取

本系列深入Pytorch官方Faster R-CNN源代码，博主会尽可能详尽地解释每一处代码，如果对你有帮助可以点点关注点点赞，有问题在评论区指出，博主会尽可能地解答。

---

Faster R-CNN论文链接
Pytorch官方Faster R-CNN的代码文档链接。

Pytorch官方使用的示例代码如下：

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)

下面主要就示例代码进行详细说明。

---

首先，初始化 Faster R-CNN 模型。

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

可以看出，这里使用的是主干网络 Resnet-50-FPN 的 Faster R-CNN。接下来 Debug 进内部代码。

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.
	构建一个主干网络为 ResNet-50-FPN 的 Faster R-CNN 模型。
	
    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.
	模型的输入应该为一个由tensors组成的列表，每个tensor的形状为[C,H,W]，对于每一个图像的元素值都应该在[0,1]的范围内，不同的图像有着不同的尺寸。
	
    The behavior of the model changes depending if it is in training or evaluation mode.
	模型有训练与评估两种模式，模型的表现取决于模型所处的模式。
	
    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
	在训练过程中，模型需要输入图像的tensor，以及目标(字典组成的列表)，其包含：
		- 边框（FloatTensor[N,4]）：真实框为[x1,y1,x2,y2]的形式，x 的值在 0~W 之间，y 的值在 0-H 之间。
		- 标签（Int64Tensor[N]）：每个真实框的类别标签。

    The model returns a ``Dict[Tensor]`` during training, containing the classification and regression
    losses for both the RPN and the R-CNN.
	在训练期间，模型返回一个 ”Dict[Tensor]“，包含 RPN 与 R-CNN 阶段的分类与回归损失。
	
    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
	在推理过程中，模型仅需要输入图像的tensor，然后返回经过后处理的预测结果以 "List[Dict[Tensor]]" 的形式，对于每一个输入图像，其 "Dict" 域如下：
		- 边框（FloatTensor[N,4]）：预测框为[x1,y1,x2,y2]的形式，x 的值在 0~W 之间，y 的值在 0~H 之间。
		- 标签（Int64Tensor[N]）：每个图像的预测标签。
		- 分数（Tensor[N]）：每个预测的分数。
	
    Faster R-CNN is exportable to ONNX for a fixed batch size with inputs images of fixed size.
	Faster R—CNN 可以被导出为一个固定批大小域固定尺寸输入图像的 ONNX 格式。
	
    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.
    参数：
    	pretrianed（bool）：如果为真，返回一个在 COCO train2017 上的预训练模型。
    	progress（bool）：如果为真，将下载进度条展示在屏幕。
    	pretrained_backbone（bool）：如果为真，返回一个在 Imagenet 上的主干网络预训练模型。
    	num_classes（int）：模型输出的种类数量（包括背景）。
    	trainable_backbone_layers（int）：从最后一个块开始可训练 ResNet 层的数量（未被冻结）。合法的值在 0~5 之间，5 意味着所有主干网络的层都是可训练的。
	"""
	# 使用 assert 判断 trainable_backbone_layers 的值是否合法
    assert trainable_backbone_layers <= 5 and trainable_backbone_layers >= 0 
    # dont freeze any layers if pretrained model or backbone is not used
    # 如果预训练模型或者预训练主干网络未被使用，不要冻结任何层。
    if not (pretrained or pretrained_backbone):
        trainable_backbone_layers = 5
    if pretrained:
        # no need to download the backbone if pretrained is set
        # 如果预训练模型被使用，就不需要下载预训练主干网络
        pretrained_backbone = False
   	# 获取 ResNet_FPN 主干网络
    backbone = resnet_fpn_backbone('resnet50', pretrained_backbone, trainable_layers=trainable_backbone_layers)
    # 获取 Faster R-CNN 模型
    model = FasterRCNN(backbone, num_classes, **kwargs)
    if pretrained:
        # 如果使用预训练模型，就下载相关的预训练模型配置
        state_dict = load_state_dict_from_url(model_urls['fasterrcnn_resnet50_fpn_coco'],
                                              progress=progress)
        # 加载模型配置到模型中
        model.load_state_dict(state_dict)
    return model # 返回模型

Debug 进获取 ResNet_FPN 主干网络对应代码。

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.
    构建一个在顶端加入FPN的ResNet主干网络。冻结主干网络中指定数量的层。

    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.
    参数：
        backbone_name (string)：resnet 架构。可能的值为 'ResNet','resnet18','resnet34','resnet50','resnet101','resnet152',
            'resnext50_32x4d','resnet101_32x8d','wide_resnet50_2','wide_resnet101_2'
        norm_layer (torchivision.ops)：建议使用默认值。相关细节请访问：
            (https://github.com/facebookresearch/maskrcnn-benchmark/issues/267)
        pretrained (bool)：如果为真，返回一个在 Imagenet 上的预训练主干网络模型
        trainable_layers (int)：从最后一个块开始可训练 ResNet 层的数量（未被冻结）。合法的值在 0~5 之间，5 意味着所有主干网络的层都是可训练的。
    """
    backbone = resnet.__dict__[backbone_name](
        pretrained=pretrained,
        norm_layer=norm_layer) # 获取resnet-50主干网络

    # select layers that wont be frozen
    # 选择被冻结的层（不参与训练）
    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
    # 仅仅当预训练主干网络被使用才冻结层
    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进获取 ResNet-50 主干网络的代码。

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 的网络结构如下图所示。

ResNet-50 采用了 BottleNeck 结构，其比 BasicNeck 更省参数。代码如下：

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

最后我们获得到 ResNet-50 的网络结构：

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)
)

之后还加如了FPN。

如上图所示，FPN 可以建立各种尺度都是强语义的特征金字塔，具体原理可以看这篇博客。FPN 在这里获取 ResNet-50 每个阶段提取的特征图加上最大池化最后一层特征图共五层特征图。其代码如下：

    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 卷积减少通道数量至256
            feat_shape = inner_lateral.shape[-2:]
            inner_top_down = F.interpolate(last_inner, size=feat_shape, mode="nearest") # 上采样
            last_inner = inner_lateral + inner_top_down # 横向连接
            results.insert(0, self.get_result_from_layer_blocks(last_inner, idx)) # 3x3 卷积消除混叠效应

        if self.extra_blocks is not None:
            results, names = self.extra_blocks(results, x, names) # 最大池化获得第五层特征图

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

        return out

至此，特征图的提取已全部完成，下面将进行 RPN(感兴趣区域的生成)。