《Very Deep Convolutional Networks for Large-Scale Image Recogition》
来源:互联网 发布:mac pro忘记登录密码 编辑:程序博客网 时间:2024/06/05 02:32
Architecture
固定输入224*224,预处理为减均值。
本文没有使用LRN,理由是没有提升网络的表现,同时增加计算和储存负担。
Configurations
卷积层的深度在每一个max-pooling之后都增加一倍,从第一层的64到之后一层的512。
Discussion
相比于AlexNet(11*11)和ZFNet(7*7)大的卷积核,本文使用更小的卷积核(3*3),理由:
一:
receptive field计算公式:(outsize-1)*stride+ksize
one 3*3 conv:3*3
two 3*3 conv:(3-1)*1+3=5
three 3*3 conv:(5-1)*1+3=7
三层3*3的conv的receptive field == 一层7*7的conv的recptive field
二:
比一个7*7的conv,三个3*3的conv多了2个非线性层,增加了网络的非线性表达能力。
三:
减小了网络的参数,假设网络的channel有C个,三个3*3的conv的参数为
1*1的conv的使用,让网络在不增加receptive field的前提下,引入了非线性,同时有升维/降维的功能(本文并没有)。
阅读全文
0 0
- 《Very Deep Convolutional Networks for Large-Scale Image Recogition》
- Very Deep Convolutional Networks for Large-Scale Image Recognition
- Very Deep Convolutional Networks for Large-Scale Image Recognition(精读)
- Very Deep Convolutional Networks for Large-Scale Image Recognition
- Very Deep Convolutional Networks for Large-Scale Image Recognition
- very deep convolutional networks for large-scale image recognition---vggnet
- VERY DEEP CONVOLUTIONAL NETWORKS FOR LARGE-SCALE IMAGE RECOGNITION
- 论文Very Deep Convolutional Networks for Large-Scale Image Recognition
- Very Deep Convolutional Networks for Large-Scale Image Classification
- Very deep convolutional networks for large-scale image recognition
- Very Deep Convolutional Networks for Large-Scale Image Recognition
- VGG - Very Deep Convolutional Networks for Large-Scale Image Recognition
- VERY DEEP CONVOLUTIONAL NETWORKS FOR LARGE-SCALE IMAGE RECOGNITION
- VERY DEEP CONVOLUTIONAL NETWORKS FOR LARGE-SCALE IMAGE RECOGNITION
- VGG--VERY DEEP CONVOLUTIONAL NETWORKS FOR LARGE-SCALE IMAGE RECOGNITION
- [深度学习论文笔记][Image Classification] Very Deep Convolutional Networks for Large-Scale Image Recognitio
- 深度学习研究理解10:Very Deep Convolutional Networks for Large-Scale Image Recognition
- 深度学习研究理解:Very Deep Convolutional Networks for Large-Scale Image Recognition
- mybatis学习之路----#{}, ${}两种传参数方式的区别--附源码解读
- Ubuntu 安装 eclipse
- 二维数组中的查找
- 正则
- 栈溢出笔记1.7 地址问题(2)
- 《Very Deep Convolutional Networks for Large-Scale Image Recogition》
- PAT 1103. Integer Factorization (30) DFS+树状数组。还是人家的代码好
- ssm 框架整合学习记录
- 栈溢出笔记1.8 字符串问题
- EBS杂项出库事务处理
- Android-签到日历
- 如何判断整型算数运算是否溢出
- 浮点数部分输出方面
- 使用ideaIU创建web工程图文说明