FPN

论文：Feature Pyramid Networks for Object Detection（CVPR 2017）速达>>

动机

浅层特征语义信息较少，但目标位置信息准确；深层特征语义信息丰富，但目标位置信息准确，探寻一种特征融合的方式，以提高检测器精度

特征金字塔

• （a）图像金字塔，生成不同尺度图像，得到对应不同尺度的特征，在测试时候较为常用

• （b）仅采用主网络最后一层（或者某一层）的特征，典型如：Faster RCNN

• （c）直接将不同尺度的特征进行融合（SSD）

• （d）顶层特征通过上采样和低层特征融合，并在每层都给出预测

Faster RCNN只用了主网络最后一层输出的Feature map，而FPN则不同：

• 对Feature map上采样
• 横向特征融合，低层次特征包含了更精确的位置信息（$1\times 1$卷积：调整通道数；起缓冲的作用，防止梯度对主干网络过大；特征组合，融合后用$3\times 3$卷积消除混叠效应）
• 各尺度Feature map独立预测
  

ResNet FPN结构实现

import torch.nn as nn
import torch.nn.functional as F
import math


__all__=['FPN']

class Bottleneck(nn.Module):
    expansion = 4

    def __init__(self, in_planes, planes, stride=1, downsample=None):
        super(Bottleneck, self).__init__()
        self.conv1 = nn.Conv2d(in_planes, 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, self.expansion * planes, kernel_size=1, bias=False)
        self.bn3 = nn.BatchNorm2d(self.expansion * planes)
        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)
        
        return out


class FPN(nn.Module):
    def __init__(self, block, layers):
        super(FPN, self).__init__()
        self.inplanes = 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)
        self.maxpool = nn.MaxPool2d(kernel_size=3, stride=2, padding=1)
        # Bottom-up layers
        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)

        # Top layer
        self.toplayer = nn.Conv2d(2048, 256, kernel_size=1, stride=1, padding=0)  # Reduce channels

        # Smooth layers
        self.smooth1 = nn.Conv2d(256, 256, kernel_size=3, stride=1, padding=1)
        self.smooth2 = nn.Conv2d(256, 256, kernel_size=3, stride=1, padding=1)
        self.smooth3 = nn.Conv2d(256, 256, kernel_size=3, stride=1, padding=1)

        # Lateral layers
        self.latlayer1 = nn.Conv2d(1024, 256, kernel_size=1, stride=1, padding=0)
        self.latlayer2 = nn.Conv2d( 512, 256, kernel_size=1, stride=1, padding=0)
        self.latlayer3 = nn.Conv2d( 256, 256, kernel_size=1, stride=1, padding=0)
        
        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 != block.expansion * planes:
            downsample  = nn.Sequential(
                nn.Conv2d(self.inplanes, block.expansion * planes, kernel_size=1, stride=stride, bias=False),
                nn.BatchNorm2d(block.expansion * planes)
            )
        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 _upsample_add(self, x, y):
        _,_,H,W = y.size()
        return F.upsample(x, size=(H,W), mode='bilinear') + y

    def forward(self, x):
        # Bottom-up
        x = self.conv1(x)
        x = self.bn1(x)
        x = self.relu(x)
        c1 = self.maxpool(x)
        
        c2 = self.layer1(c1)
        c3 = self.layer2(c2)
        c4 = self.layer3(c3)
        c5 = self.layer4(c4)
        # Top-down
        p5 = self.toplayer(c5)
        p4 = self._upsample_add(p5, self.latlayer1(c4))
        p3 = self._upsample_add(p4, self.latlayer2(c3))
        p2 = self._upsample_add(p3, self.latlayer3(c2))
        # Smooth
        p4 = self.smooth1(p4)
        p3 = self.smooth2(p3)
        p2 = self.smooth3(p2)
        return p2, p3, p4, p5


def FPN101():
    return FPN(Bottleneck, [2,2,2,2])

def test():
    net = FPN101()
    fms = net(Variable(torch.randn(1,3,600,900)))
    for fm in fms:
        print(fm.size())

test()
FPN101()
>>>
torch.Size([1, 256, 150, 225])
torch.Size([1, 256, 75, 113])
torch.Size([1, 256, 38, 57])
torch.Size([1, 256, 19, 29])
FPN(
  (conv1): Conv2d(3, 64, kernel_size=(7, 7), stride=(2, 2), padding=(3, 3), bias=False)
  (bn1): BatchNorm2d(64, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
  (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): BatchNorm2d(64, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
      (conv2): Conv2d(64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)
      (bn2): BatchNorm2d(64, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
      (conv3): Conv2d(64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False)
      (bn3): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
      (relu): ReLU(inplace=True)
      (downsample): Sequential(
        (0): Conv2d(64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False)
        (1): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
      )
    )
    (1): Bottleneck(
      (conv1): Conv2d(256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False)
      (bn1): BatchNorm2d(64, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
      (conv2): Conv2d(64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)
      (bn2): BatchNorm2d(64, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
      (conv3): Conv2d(64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False)
      (bn3): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
      (relu): ReLU(inplace=True)
    )
  )
  (layer2): Sequential(
    (0): Bottleneck(
      (conv1): Conv2d(256, 128, kernel_size=(1, 1), stride=(1, 1), bias=False)
      (bn1): BatchNorm2d(128, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
      (conv2): Conv2d(128, 128, kernel_size=(3, 3), stride=(2, 2), padding=(1, 1), bias=False)
      (bn2): BatchNorm2d(128, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
      (conv3): Conv2d(128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False)
      (bn3): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
      (relu): ReLU(inplace=True)
      (downsample): Sequential(
        (0): Conv2d(256, 512, kernel_size=(1, 1), stride=(2, 2), bias=False)
        (1): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
      )
    )
    (1): Bottleneck(
      (conv1): Conv2d(512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False)
      (bn1): BatchNorm2d(128, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
      (conv2): Conv2d(128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)
      (bn2): BatchNorm2d(128, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
      (conv3): Conv2d(128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False)
      (bn3): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
      (relu): ReLU(inplace=True)
    )
  )
  (layer3): Sequential(
    (0): Bottleneck(
      (conv1): Conv2d(512, 256, kernel_size=(1, 1), stride=(1, 1), bias=False)
      (bn1): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
      (conv2): Conv2d(256, 256, kernel_size=(3, 3), stride=(2, 2), padding=(1, 1), bias=False)
      (bn2): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
      (conv3): Conv2d(256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False)
      (bn3): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
      (relu): ReLU(inplace=True)
      (downsample): Sequential(
        (0): Conv2d(512, 1024, kernel_size=(1, 1), stride=(2, 2), bias=False)
        (1): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
      )
    )
    (1): Bottleneck(
      (conv1): Conv2d(1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False)
      (bn1): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
      (conv2): Conv2d(256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)
      (bn2): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
      (conv3): Conv2d(256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False)
      (bn3): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
      (relu): ReLU(inplace=True)
    )
  )
  (layer4): Sequential(
    (0): Bottleneck(
      (conv1): Conv2d(1024, 512, kernel_size=(1, 1), stride=(1, 1), bias=False)
      (bn1): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
      (conv2): Conv2d(512, 512, kernel_size=(3, 3), stride=(2, 2), padding=(1, 1), bias=False)
      (bn2): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
      (conv3): Conv2d(512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False)
      (bn3): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
      (relu): ReLU(inplace=True)
      (downsample): Sequential(
        (0): Conv2d(1024, 2048, kernel_size=(1, 1), stride=(2, 2), bias=False)
        (1): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
      )
    )
    (1): Bottleneck(
      (conv1): Conv2d(2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False)
      (bn1): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
      (conv2): Conv2d(512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)
      (bn2): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
      (conv3): Conv2d(512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False)
      (bn3): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
      (relu): ReLU(inplace=True)
    )
  )
  (toplayer): Conv2d(2048, 256, kernel_size=(1, 1), stride=(1, 1))
  (smooth1): Conv2d(256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
  (smooth2): Conv2d(256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
  (smooth3): Conv2d(256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
  (latlayer1): Conv2d(1024, 256, kernel_size=(1, 1), stride=(1, 1))
  (latlayer2): Conv2d(512, 256, kernel_size=(1, 1), stride=(1, 1))
  (latlayer3): Conv2d(256, 256, kernel_size=(1, 1), stride=(1, 1))
)

FPN中的RoI 与 Anchor

RoI 与特征层

与RPN结合使用时，根据RPN预测ROIs的尺寸截取对应特征层作为ROI pooling层的输入，大尺寸ROI用相对较后的特征层（如：P5）；小尺寸ROI就用靠前的特征层（P4、P3等），判断依据：
$$k=\lfloor k_0+lo{g}_2(\sqrt{wh}/224)\rfloor$$

其中 224 是ImageNet的标准输入，$k_0$ 为基准值（设为5，代表$P_5$特征层），$w、h$为ROI宽高（例如：ROI：$w\times h=112\times 112$，则$k=K_0+lo{g}_2\frac{1}{2}=5-1=4⟹$用$P_4$特征层）。

roi_level = min(5, max(2, 4 + log2(sqrt(w * h) / ( 224 / sqrt(image_area)))))

Anchor 与特征层

RPN网络以单个尺度的feature map作为输入。而FPN可以生成不同尺度特征作为RPN网络的输入。每个特征层上都定义了不同大小的Anchor，$P_2，P_3，P_4，P_5，P_6$分别对应面积大小为$32^2,64^2,128^2,256^2,512^2$的Anchor，浅层对应小面积的小目标。为适应不同形状，每层的Anchor依然设置3种长宽比例：$[1:2，1:1，2:1]$，共15种 anchor

实验

参考文献

【1】FPN详解
【2】FPN（feature pyramid networks）算法讲解
【3】FPN: 一种高效的CNN特征提取方法
【4】RetinaNet: Focal loss在目标检测中的应用
【5】FPN（特征图金字塔网络）理论基础与具体实现
【6】FPN（特征金字塔网络）的直觉、架构和表现简要介绍
【7】Feature Pyramid Networks for Object Detection 论文笔记
【8】FPN提到的混叠效应用3x3的卷积核消除，这里的混叠效应应该作何理解？
【9】从代码细节理解 FPN （Feature Pyramid Networks for Object Detection）
【10】FAIR最新视觉论文集锦：FPN，RetinaNet，Mask和Mask-X RCNN（含代码实现）