在CNN(1)和CNN(2)两篇文章中,主要说明的是CNN的基本架构和权值共享(Weight Sharing),本文则重点介绍卷积的部分。

首先,在卷积之前,我们的数据是4D的tensor(width,height,channels,batch),在CNN(1):Architecture一文中,曾经提到过。这里的通道,和之前所说的depth是一个概念,例如一张Grey Scale Image,其通道数为1;RGB图的通道数为3。

而事实上,kernel也是有channel的,并且其数量与输入tensor的通道数一致。如下图的RGB图像,有3个通道,在卷积过程中我们使用5个kernel。

那么,到底有多少feature map?有很长一段时间,我的答案是15个feature map。但实际上,答案是5个(和kernel的数量一致)。如图示,Kernel也有3个通道,分别与RGB三个通道进行卷积后,生成了3个单通道卷积结果,而后三个结果相加得到feature map。而在此卷积层中,kernel是我们唯独需要“学习”的参数,也就是3*3*3*5=45个参数。实际上,通过训练kernel会学习到应该过滤图片中的哪些特征。

Convolutional Neural Networks(3):Convolution and Channels的更多相关文章

  1. [CS231n-CNN] Convolutional Neural Networks: architectures, convolution / pooling layers

    课程主页:http://cs231n.stanford.edu/     参考: 细说卷积神经网络:http://blog.csdn.net/han_xiaoyang/article/details/ ...

  2. [转] Understanding Convolutional Neural Networks for NLP

    http://www.wildml.com/2015/11/understanding-convolutional-neural-networks-for-nlp/ 讲CNN以及其在NLP的应用,非常 ...

  3. 课程四(Convolutional Neural Networks),第一周(Foundations of Convolutional Neural Networks) —— 3.Programming assignments:Convolutional Model: application

    Convolutional Neural Networks: Application Welcome to Course 4's second assignment! In this notebook ...

  4. 课程四(Convolutional Neural Networks),第一周(Foundations of Convolutional Neural Networks) —— 2.Programming assignments:Convolutional Model: step by step

    Convolutional Neural Networks: Step by Step Welcome to Course 4's first assignment! In this assignme ...

  5. Convolutional Neural Networks: Step by Step

    Andrew Ng deeplearning courese-4:Convolutional Neural Network Convolutional Neural Networks: Step by ...

  6. Convolutional Neural Networks: Application

    Andrew Ng deeplearning courese-4:Convolutional Neural Network Convolutional Neural Networks: Step by ...

  7. (zhuan) Building Convolutional Neural Networks with Tensorflow

    Ahmet Taspinar Home About Contact Building Convolutional Neural Networks with Tensorflow Posted on a ...

  8. [转]An Intuitive Explanation of Convolutional Neural Networks

    An Intuitive Explanation of Convolutional Neural Networks https://ujjwalkarn.me/2016/08/11/intuitive ...

  9. Understanding Convolutional Neural Networks for NLP

    When we hear about Convolutional Neural Network (CNNs), we typically think of Computer Vision. CNNs ...

随机推荐

  1. CentOS 5.5编译安装lnmp

    如果是安装Centos6.5记得Perl是必选的,否则无法安装VMWare Tools!!!!切记 如果出现make错误需要安装其他软件,装好后  make clean   make install ...

  2. XE2再次安装JVCL

    说再次安装,是因为我忘了自己之前写过的方法,装完之后感觉怪怪的,想看看别人是怎么装的,于是用google搜索了一下,才发现之前写过: http://www.cnblogs.com/findumars/ ...

  3. SQL语句-exec执行

    动态sql语句基本语法1 :普通SQL语句可以用Exec执行 eg:   Select * fromtableName        Exec('select * from tableName')   ...

  4. spark复习笔记(3):使用spark实现单词统计

    wordcount是spark入门级的demo,不难但是很有趣.接下来我用命令行.scala.Java和python这三种语言来实现单词统计. 一.使用命令行实现单词的统计 1.首先touch一个a. ...

  5. Webstorm上已有的本地项目上传到Github

    1.Webstorm左上角File —— Settings—— Github, Auth Type改成password,在Login和Password中输入自己的github账户和密码,点Test测试 ...

  6. Git同步问题

    1. 无法合并不相关历史记录 git pull origin master --allow-unrelated-histories

  7. 行人重识别(ReID) ——数据集描述 Market-1501

    数据集简介 Market-1501 数据集在清华大学校园中采集,夏天拍摄,在 2015 年构建并公开.它包括由6个摄像头(其中5个高清摄像头和1个低清摄像头)拍摄到的 1501 个行人.32668 个 ...

  8. 【记录】@Transactional

    参考链接:https://blog.csdn.net/nextyu/article/details/78669997 参考链接:https://www.xuebuyuan.com/3222458.ht ...

  9. break语句、continue语句、goto语句的用法辨析

    1.break语句 break语句常使用在switch语句.循环体以及if语句中,它的作用是跳出循环,而且只能跳出一层循环. for (i = 0; i < 10; ++j) { for (j ...

  10. ShellExecute指定IE浏览器打开

    ShellExecute(NULL,L"open", L"iexplore.exe", L"www.baidu.com", NULL, SW ...