关闭

Convolutional Networks

标签: CNN
235人阅读 评论(0) 收藏 举报
分类:
 CNN

1. Overall architecture
It is my first time to learn the CNN, which is hot field in machine learning. I will try to speak detail of CNN.
convolutional network

> Q1: what is the channel?
The channel, which is the number of the input for special layer, could be image or feature map. For example(1), The number of channel is three in the first layer of CNN if the network is used to deal with RGB picture. (2) In the hidden layers, if the number of output of K layer is 32, then the number of channel of K+1 layer is previous output of K layer, that is 32.
> How the feature maps in k layer are performed by convolution layer?
convolutions
When performing the input features, the convolutional network deal with the patches from the same area of all previous layers.
> what is the convolution kernel?
Generally, the convolution kernel is be selected from the point that it is similar to the mechanism of brain which recognize the object feature. Based on the analysis, the Sobel or Laplace operator are used as convolution kernel.
the Sobel operator
Laplace
>what is the parameter sharing
Units organized into the same “feature map” share parameters
parameter sharing
In experiment, the distribute of parameter follows the Guussian or Xarvier distribution.
> what is the feature map?
For CNN, one convolution kernel means one feature map that is convolution kernel perform input data or feature maps to get certain feature map.
convolutional network
where every element of the set (64 64 256 256) is the number of feature maps in currency layer.
> what is the local connectivity
Each hidden layer is only connected with the subregion (patch) of the input image. Besides, it is connected to all channels.
local connectivity

>what is the pooling
The pooling is performed on the non-overlapping neighborhoods. However, there will has overlapping neighborhoods.
Max_Pooling
Average_Pooling
> what is the active function?
sigmoid
tanh

0
0
查看评论

Inverting Convolutional Networks with Convolutional Networks 论文理解

Inverting Convolutional Networks with Convolutional Networks  论文创新点: (1)提出了一个前向生成图像网络(上卷积网络)。相比于之前的梯度下降优化方式提高了速度,基本能达到实时。 (2)提出了新的损耗计算方式。图像...
  • wyl1987527
  • wyl1987527
  • 2017-06-14 21:11
  • 157

深度学习入门:Fully Convolutional Networks

Fully Convolutional Networks
  • liyaohhh
  • liyaohhh
  • 2016-03-05 10:14
  • 7921

CNN(Convolutional Neural Networks)没有原理只有实现

零.说明:        本文的所有代码均可在 DML 找到,欢迎点星星。         注.CNN的这份代码非常慢,基本上没有实际使用的可能,所以我只是发出来,代表我还是实践过而已 一.引入: &#...
  • Dark_Scope
  • Dark_Scope
  • 2013-12-03 23:03
  • 20135

【论文笔记】Fully Convolutional Networks for Semantic Segmentation

【Author:Jonathan Long  Evan Shelhamer Publish:CVPR】 卷积网络的贡献: 改进了全图像分类[19,31,32]在结构化输出的局部任务上取得了进展:目标检测边界框、部位和关键点预测、局部匹配(通信) 从粗略到精细推理的下一步是像素预测。 ...
  • Wonder233
  • Wonder233
  • 2017-04-25 10:25
  • 1349

Visualizing and Understanding Convolutional Networks翻译总结

Zfnet 论文:《Visualizing and Understanding Convolutional Networks》 1.     概述 这篇论文主要的贡献是将卷及神经网络的隐藏层的特征进行可视化,后面通过可视化的展示来分析如何构建更好的网络结构。最...
  • xjz18298268521
  • xjz18298268521
  • 2016-08-31 09:35
  • 2345

CNN入门必读经典:Visualizing and Understanding Convolutional Networks

本文主要是借助deconvnet来可视化卷积网络,这对于理解卷积网络还是非常重要的,同时本文又是13年ImageNet分类任务的冠军。 代码: https://github.com/guruucsd/CNN_visualization1 Deconvolution首先我们先对Deconvolut...
  • bea_tree
  • bea_tree
  • 2017-04-03 12:55
  • 1478

Visual Tracking with Fully Convolutional Networks 笔记

简单介绍一下背景,这篇文章是大连理工的卢湖川教授http://202.118.75.4/lu/publications.html的学生Lijun Wang在港中文与Xiaogang Wang团队合作的ICCV2015的文章。笔者7月份在CUHK听报告的时候有幸提前看到相关的展示,感觉结果很惊人。
  • carrierlxksuper
  • carrierlxksuper
  • 2015-10-06 10:12
  • 5403

Image Super-Resolution Using Deep Convolutional Network (with demo)

Paper:  Chao Dong, Chen Change Loy, Kaiming He, Xiaoou Tang, "Image Super-Resolution Using Deep Convolutional Networks", http:/...
  • yihaizhiyan
  • yihaizhiyan
  • 2015-08-05 11:01
  • 1592

论文阅读笔记:Fully Convolutional Networks for Semantic Segmentation

【概览&主要贡献】 【问题&解决办法】 如何进行dense prediction 如何refine,得到更好的结果 【结果】
  • tangwei2014
  • tangwei2014
  • 2015-07-14 20:07
  • 24502

文献《Densely Connected Convolutional Networks》阅读笔记

论文的主要贡献: 论文提出了一种新的网络DenseNet。    论文想法的来源: 卷积神经网络越深,准确率越高,同时与输入靠近的层和输出靠近的层时间的连接越短,越有助于训练。    论文的网络DenseNets: 是将要通过卷积神经网络的图像。网络一共包含L...
  • qq_14948653
  • qq_14948653
  • 2017-05-27 21:43
  • 1819
    个人资料
    • 访问:6313次
    • 积分:180
    • 等级:
    • 排名:千里之外
    • 原创:9篇
    • 转载:5篇
    • 译文:0篇
    • 评论:1条
    文章分类
    文章存档
    最新评论