Squeeze-and-Excitation Networks

发表于 2020-06-12 更新于 2021-07-09 分类于目标分类/object classification 阅读次数：

本文字数： 3.5k 阅读时长 ≈ 6 分钟

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.

卷积神经网络的核心构件是卷积算子，它使网络能够通过在每层的局部感受野内融合空间和通道信息来构造信息特征。大量的前期研究已经调查了这种关系的空间成分，试图通过提高整个特征层次的空间编码质量来增强CNN的表达能力。在这项工作中，我们将重点放在通道关系上，并提出了一个新的架构单元，我们称之为“挤压和激励”(SE)模块，它通过显式建模通道之间的相互依赖性，自适应地重校准逐通道的特征响应。我们表明，这些块堆叠在一起得到的SENet架构能够非常有效的泛化到不同数据集。我们进一步证明，在增加少量额外计算成本的情况下，SE模块能够给最先进CNN网络带来了显著的性能改进。挤压和激励网络构成了我们的ILSVRC 2017分类提交的基础，它赢得了第一名，将top-5误差率降低到2.251%，超过了2016年的获胜者（相对提高了25%）。模型和代码公布在 https://github.com/hujie-frank/SENet

章节内容

首先介绍挤压激励单元的实现及数学推导
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
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
通过烧蚀研究分析了SE模块的组成以及作用
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\(X\)到\(U\)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：

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

其中，\(\ast\)表示卷积操作；\(X=[x^{1}, x^{2}, ..., x^{C'}]; v_{c} =[v_{c}^{1}, v_{c}^{2}, ..., v_{c}^{C'}]\)

在完成\(F_{tr}\)操作后，对输出特征图\(U\)执行Sequeeze-and-Excitation操作