caffe中的BatchNorm层
在训练一个小的分类网络时，发现加上BatchNorm层之后的检索效果相对于之前，效果会有提升，因此将该网络结构记录在这里，供以后查阅使用：
添加该层之前：
layer {
name: "conv1"
type: "Convolution"
bottom: "data"
top: "conv1"
param {
lr_mult:
decay_mult:
}
param {
lr_mult:
decay_mult:
}
convolution_param {
num_output:
kernel_size:
stride:
weight_filler {
type: "gaussian"
std: 0.01
}
bias_filler {
type: "constant"
value:
}
}
}
layer {
name: "relu1"
type: "ReLU"
bottom: "conv1"
top: "conv1"
}
layer {
name: "pool1"
type: "Pooling"
bottom: "conv1"
top: "pool1"
pooling_param {
pool: MAX
kernel_size:
stride:
}
}
layer {
name: "norm1"
type: "LRN"
bottom: "pool1"
top: "norm1"
lrn_param {
local_size:
alpha: 0.0001
beta: 0.75
}
}
layer {
name: "conv2"
type: "Convolution"
bottom: "norm1"
top: "conv2"
param {
lr_mult:
decay_mult:
}
param {
lr_mult:
decay_mult:
}
convolution_param {
num_output:
pad:
kernel_size:
stride:
group:
weight_filler {
type: "gaussian"
std: 0.01
}
bias_filler {
type: "constant"
value:
}
}
}
layer {
name: "relu2"
type: "ReLU"
bottom: "conv2"
top: "conv2"
}
layer {
name: "pool2"
type: "Pooling"
bottom: "conv2"
top: "pool2"
pooling_param {
pool: MAX
kernel_size:
stride:
}
}
layer {
name: "norm2"
type: "LRN"
bottom: "pool2"
top: "norm2"
lrn_param {
local_size:
alpha: 0.0001
beta: 0.75
}
}
layer {
name: "conv3"
type: "Convolution"
bottom: "norm2"
top: "conv3"
param {
lr_mult:
decay_mult:
}
param {
lr_mult:
decay_mult:
}
convolution_param {
num_output:
pad:
stride:
kernel_size:
weight_filler {
type: "gaussian"
std: 0.01
}
bias_filler {
type: "constant"
value:
}
}
}
layer {
name: "relu3"
type: "ReLU"
bottom: "conv3"
top: "conv3"
}
layer {
name: "conv4"
type: "Convolution"
bottom: "conv3"
top: "conv4"
param {
lr_mult:
decay_mult:
}
param {
lr_mult:
decay_mult:
}
convolution_param {
num_output:
pad:
kernel_size:
group:
weight_filler {
type: "gaussian"
std: 0.01
}
bias_filler {
type: "constant"
value:
}
}
}
layer {
name: "relu4"
type: "ReLU"
bottom: "conv4"
top: "conv4"
}
layer {
name: "conv5"
type: "Convolution"
bottom: "conv4"
top: "conv5"
param {
lr_mult:
decay_mult:
}
param {
lr_mult:
decay_mult:
}
convolution_param {
num_output:
pad:
stride:
kernel_size:
group:
weight_filler {
type: "gaussian"
std: 0.01
}
bias_filler {
type: "constant"
value:
}
}
}
layer {
name: "relu5"
type: "ReLU"
bottom: "conv5"
top: "conv5"
}
layer {
name: "pool5"
type: "Pooling"
bottom: "conv5"
top: "pool5"
pooling_param {
pool: MAX
kernel_size:
stride:
}
}
layer {
name: "fc6_srx"
type: "InnerProduct"
bottom: "pool5"
top: "fc6"
param {
lr_mult:
decay_mult:
}
param {
lr_mult:
decay_mult:
}
inner_product_param {
num_output:
weight_filler {
type: "gaussian"
std: 0.005
}
bias_filler {
type: "constant"
value:
}
}
}
layer {
name: "relu7"
type: "ReLU"
bottom: "fc6"
top: "fc6"
}
layer {
name: "drop7"
type: "Dropout"
bottom: "fc6"
top: "fc6"
dropout_param {
dropout_ratio: 0.5
}
}
layer {
name: "fc7_srx"
type: "InnerProduct"
bottom: "fc6"
top: "fc7"
param {
lr_mult:
decay_mult:
}
param {
lr_mult:
decay_mult:
}
inner_product_param {
num_output:
weight_filler {
type: "gaussian"
std: 0.01
}
bias_filler {
type: "constant"
value:
}
}
}
layer{
name: "loss"
type: "SoftmaxWithLoss"
top: "SoftmaxWithLoss"
bottom: "fc7"
bottom: "label"
include {
phase: TRAIN
}
}
layer {
name: "accuracy"
type: "Accuracy"
bottom: "fc7"
bottom: "label"
top: "accuracy"
include {
phase: TEST
}
}
添加该层之后：
layer {
name: "conv1"
type: "Convolution"
bottom: "data"
top: "conv1"
param {
lr_mult:
decay_mult:
}
param {
lr_mult:
decay_mult:
}
convolution_param {
num_output:
kernel_size:
stride:
weight_filler {
type: "gaussian"
std: 0.01
}
bias_filler {
type: "constant"
value:
}
}
}
##############
layer {
bottom: "conv1"
top: "conv1"
name: "bn1"
type: "BatchNorm"
param {
lr_mult:
}
param {
lr_mult:
}
param {
lr_mult:
}
}
##############
layer {
name: "relu1"
type: "ReLU"
bottom: "conv1"
top: "conv1"
}
layer {
name: "pool1"
type: "Pooling"
bottom: "conv1"
top: "pool1"
pooling_param {
pool: MAX
kernel_size:
stride:
}
}
layer {
name: "conv2"
type: "Convolution"
bottom: "pool1"
top: "conv2"
param {
lr_mult:
decay_mult:
}
param {
lr_mult:
decay_mult:
}
convolution_param {
num_output:
pad:
kernel_size:
stride:
group:
weight_filler {
type: "gaussian"
std: 0.01
}
bias_filler {
type: "constant"
value:
}
}
}
##############
layer {
bottom: "conv2"
top: "conv2"
name: "bn2"
type: "BatchNorm"
param {
lr_mult:
}
param {
lr_mult:
}
param {
lr_mult:
}
}
##############
layer {
name: "relu2"
type: "ReLU"
bottom: "conv2"
top: "conv2"
}
layer {
name: "pool2"
type: "Pooling"
bottom: "conv2"
top: "pool2"
pooling_param {
pool: MAX
kernel_size:
stride:
}
}
layer {
name: "conv3"
type: "Convolution"
bottom: "pool2"
top: "conv3"
param {
lr_mult:
decay_mult:
}
param {
lr_mult:
decay_mult:
}
convolution_param {
num_output:
pad:
stride:
kernel_size:
weight_filler {
type: "gaussian"
std: 0.01
}
bias_filler {
type: "constant"
value:
}
}
}
##############
layer {
bottom: "conv3"
top: "conv3"
name: "bn3"
type: "BatchNorm"
param {
lr_mult:
}
param {
lr_mult:
}
param {
lr_mult:
}
}
##############
layer {
name: "relu3"
type: "ReLU"
bottom: "conv3"
top: "conv3"
}
layer {
name: "conv4"
type: "Convolution"
bottom: "conv3"
top: "conv4"
param {
lr_mult:
decay_mult:
}
param {
lr_mult:
decay_mult:
}
convolution_param {
num_output:
pad:
kernel_size:
group:
weight_filler {
type: "gaussian"
std: 0.01
}
bias_filler {
type: "constant"
value:
}
}
}
##############
layer {
bottom: "conv4"
top: "conv4"
name: "bn4"
type: "BatchNorm"
param {
lr_mult:
}
param {
lr_mult:
}
param {
lr_mult:
}
}
##############
layer {
name: "relu4"
type: "ReLU"
bottom: "conv4"
top: "conv4"
}
layer {
name: "conv5"
type: "Convolution"
bottom: "conv4"
top: "conv5"
param {
lr_mult:
decay_mult:
}
param {
lr_mult:
decay_mult:
}
convolution_param {
num_output:
pad:
stride:
kernel_size:
group:
weight_filler {
type: "gaussian"
std: 0.01
}
bias_filler {
type: "constant"
value:
}
}
}
##############
layer {
bottom: "conv5"
top: "conv5"
name: "bn5"
type: "BatchNorm"
param {
lr_mult:
}
param {
lr_mult:
}
param {
lr_mult:
}
}
##############
layer {
name: "relu5"
type: "ReLU"
bottom: "conv5"
top: "conv5"
}
layer {
name: "pool5"
type: "Pooling"
bottom: "conv5"
top: "pool5"
pooling_param {
pool: MAX
kernel_size:
stride:
}
}
layer {
name: "fc6_srx"
type: "InnerProduct"
bottom: "pool5"
top: "fc6"
param {
lr_mult:
decay_mult:
}
param {
lr_mult:
decay_mult:
}
inner_product_param {
num_output:
weight_filler {
type: "gaussian"
std: 0.005
}
bias_filler {
type: "constant"
value:
}
}
}
layer {
name: "relu7"
type: "ReLU"
bottom: "fc6"
top: "fc6"
}
layer {
name: "drop7"
type: "Dropout"
bottom: "fc6"
top: "fc6"
dropout_param {
dropout_ratio: 0.5
}
}
layer {
name: "fc7_srx"
type: "InnerProduct"
bottom: "fc6"
top: "fc7"
param {
lr_mult:
decay_mult:
}
param {
lr_mult:
decay_mult:
}
inner_product_param {
num_output:
weight_filler {
type: "gaussian"
std: 0.01
}
bias_filler {
type: "constant"
value:
}
}
}
layer{
name: "loss"
type: "SoftmaxWithLoss"
top: "SoftmaxWithLoss"
bottom: "fc7"
bottom: "label"
include {
phase: TRAIN
}
}
layer {
name: "accuracy"
type: "Accuracy"
bottom: "fc7"
bottom: "label"
top: "accuracy"
include {
phase: TEST
}
}
caffe中的BatchNorm层的更多相关文章
- （原）torch和caffe中的BatchNorm层
转载请注明出处: http://www.cnblogs.com/darkknightzh/p/6015990.html BatchNorm具体网上搜索. caffe中batchNorm层是通过Batc ...
- caffe 中 python 数据层
caffe中大多数层用C++写成. 但是对于自己数据的输入要写对应的输入层,比如你要去图像中的一部分,不能用LMDB,或者你的label 需要特殊的标记. 这时候就需要用python 写一个输入层. ...
- caffe中全卷积层和全连接层训练参数如何确定
今天来仔细讲一下卷基层和全连接层训练参数个数如何确定的问题.我们以Mnist为例,首先贴出网络配置文件: name: "LeNet" layer { name: "mni ...
- caffe中添加local层
下载caffe-local,解压缩; 修改makefile.config:我是将cuudn注释掉,去掉cpu_only的注释; make all make test(其中local_test出错,将文 ...
- 转载：caffe中的Reshape层
http://blog.csdn.net/terrenceyuu/article/details/76228317 #作用:在不改变数据的情况下,改变输入的维度 layer { name: " ...
- caffe中各层的作用:
关于caffe中的solver: cafffe中的sover的方法都有: Stochastic Gradient Descent (type: "SGD"), AdaDelta ( ...
- 深度学习中 batchnorm 层是咋回事？
作者:Double_V_ 来源:CSDN 原文:https://blog.csdn.net/qq_25737169/article/details/79048516 版权声明:本文为博主原创文章,转载 ...
- caffe中ConvolutionLayer的前向和反向传播解析及源码阅读
一.前向传播 在caffe中,卷积层做卷积的过程被转化成了由卷积核的参数组成的权重矩阵weights(简记为W)和feature map中的元素组成的输入矩阵(简记为Cin)的矩阵乘积W * Cin. ...
- caffe中batch norm源码阅读
1. batch norm 输入batch norm层的数据为[N, C, H, W], 该层计算得到均值为C个,方差为C个,输出数据为[N, C, H, W]. <1> 形象点说,均值的 ...
随机推荐
- Python~删除空格，插入换行符号
f.write(rf.replace(' ','')) f.write(rf.replace('1041','\n1041')) 不能连续起作用? # -*- coding: UTF-8 -*- im ...
- iOS 归档archive使用方法
归档是一种很常用的文件储存方法,几乎任何类型的对象都能够被归档储存,文件将被保存成自定 义类型的文件,相对于NSUserDefault具有更好的保密性. 1.使用archiveRootObject ...
- NEFU 558 迷宫寻路
题目链接 简单搜索题 #include <cstdio> #include <iostream> #include <cstring> using namespac ...
- jquery ashx交互 返回list 循环json输入信息
html代码:触发按钮 <input type="button" id="search" value="查询" /> ashx代 ...
- linq 的switch实现
List<RemindSend> lrs = (from a in db.Remind join b in db.Certified on a.certifiedId equals b.C ...
- C语言字符串操作总结大全
1)字符串操作 strcpy(p, p1) 复制字符串 函数原型strncpy(p, p1, n) 复制指定长度字符串 函数原型strcat(p, p1) 附加字符串 函数原型strn ...
- 第一章 Web应用程序简介
1.1 Web应用程序基础知识 URL.URN与URI URL:Uniform Resource Locator URN:Uniform Resource Name URI:Uniform Resou ...
- 复利计算--结对项目<04-11-2016> 1.0.0 lastest 阶段性完工~
结对项目:Web复利计算 搭档博客地址:25江志彬 http://www.cnblogs.com/qazwsxedcrfv/ 个人摘要: (2016-04-09-12:00)补充:之前传送门没做好, ...
- Python lambda函数
python允许定义单行的小函数,定义lambda函数的形式如下: lambda 参数:表达式lambda函数默认返回表达式的值,可接收任意个参数,包括可选参数,但是表达式只有一个.
- hasLayout && Block Formatting Contexts
转自:http://www.smallni.com/haslayout-block-formatting-contexts/ 因为本人脑子不好使,自己打印出了一张hasLayout和Block For ...