## 摘要

The central building block of convolutional neural networks (CNNs) is the convolution operator, which enables networks to construct informative features by fusing both spatial and channel-wise information within local receptive fields at each layer. A broad range of prior research has investigated the spatial component of this relationship, seeking to strengthen the representational power of a CNN by enhancing the quality of spatial encodings throughout its feature hierarchy. In this work, we focus instead on the channel relationship and propose a novel architectural unit, which we term the “Squeeze-and-Excitation” (SE) block, that adaptively recalibrates channel-wise feature responses by explicitly modelling interdependencies between channels. We show that these blocks can be stacked together to form SENet architectures that generalise extremely effectively across different datasets. We further demonstrate that SE blocks bring significant improvements in performance for existing state-of-the-art CNNs at slight additional computational cost. Squeeze-and-Excitation Networks formed the foundation of our ILSVRC 2017 classification submission which won first place and reduced the top-5 error to 2.251%, surpassing the winning entry of 2016 by a relative improvement of ∼25%. Models and code are available at https://github.com/hujie-frank/SENet.

## 章节内容

1. 首先介绍挤压激励单元的实现及数学推导
2. 其次比较了嵌入SE模块的模型和原始模型之间的大小和计算复杂度
3. 接着通过实验证明了SE模块在不同任务（目标识别、检测等）、不同架构（Inception/ResNet等）、不同深度、不同数据集中的泛化能力
4. 通过烧蚀研究分析了SE模块的组成以及作用
5. 通过实验证明了Squeeze操作和Excitation操作的不可或缺

## SE模块

Sequeeze-and-Excitation block的实现如上图所示。$$X$$表示输入数据，大小为$${H}'\times {W}'\times {C}'$$$$U$$表示特征图，大小为$$H\times W\times C$$$$F_{tr}$$表示$$X$$$$U$$之间的某种转换。假设$$F_{tr}$$是一个卷积操作，$$V=[v_{1}, v_{2}, ..., v_{C}]$$表示滤波器，那么计算如下：

$u_{c}=v_{c} \ast X =\sum_{s=1}^{C'} v_{c}^{s} \ast X^{s}$

### Squeeze单元

$z_{c}=F_{sq}(u_{c}) = \frac {1}{H\times W}\sum_{i=1}^{H}\sum_{j=1}^{W}u_{c}(i,j)$

### Excitation单元

$s=F_{ex}(z, W) = \sigma(g(z, W)) = \sigma(W_{2}\delta(W_{1}z))$

• $$\delta = ReLU$$
• $$\sigma = Sigmoid$$
• $$W_{1}\in R^{\frac {C}{r}\times C}$$
• $$W_{2}\in R^{C\times \frac {C}{r}}$$

### Scale

Excitation单元得到自适应重校准后的全局信息后，对输出特征$$U$$进行特征缩放

$\tilde{x}_{c} = F_{scale}(u_{c}, s_{c}) = s_{c}u_{c}$

• $$\tilde{X} =[\tilde{x}_{1}, \tilde{x}_{2}, ..., \tilde{x}_{C}]$$

### 小结

SE模块可以适用于任意的卷积操作之后，论文还给出了SE-InceptionSE-ResNet模块的实现

## 烧蚀研究

### Excitation算子

Excitation实现中，后一个全连接层使用了Sigmoid激活函数，论文还比较了ReLUtanh

### 嵌入策略

1. SE-PRE：在残差单元之前嵌入
2. SE-POST：在一致性映射（求和+ReLU）之后嵌入
3. SE-IdentitySE单元嵌入到一致性连接分支，并行于残差单元