CS231n课程笔记:Leture5 Convolutional Neural Networks

目录

convolution                                                                                                                    

Pooling


CS231n课程笔记:Leture5 Convolutional Neural Networks_第1张图片

   convolutionCS231n课程笔记:Leture5 Convolutional Neural Networks_第2张图片                                                                                                                                                            CS231n课程笔记:Leture5 Convolutional Neural Networks_第3张图片       how  to visualize these features?                                                                                                                                                            CS231n课程笔记:Leture5 Convolutional Neural Networks_第4张图片

 CS231n课程笔记:Leture5 Convolutional Neural Networks_第5张图片

 if stride 2

CS231n课程笔记:Leture5 Convolutional Neural Networks_第6张图片

 what about stride 3?

CS231n课程笔记:Leture5 Convolutional Neural Networks_第7张图片

 zero pad the borders

CS231n课程笔记:Leture5 Convolutional Neural Networks_第8张图片

 padding value *2 + N   

If you got any questions here,you can learn from this blog in details(21条消息) CNN中卷积核数和输出通道数的关系_co2e的博客-CSDN博客_卷积核个数与输入输出通道数关系

ok Question:

Input volume :32x32x3

10 5x5 filters with stride 1, pad 2

number of parameters in this layer?

bias! 

CS231n课程笔记:Leture5 Convolutional Neural Networks_第9张图片\CS231n课程笔记:Leture5 Convolutional Neural Networks_第10张图片

 example: conv layer in torch

 Pooling

CS231n课程笔记:Leture5 Convolutional Neural Networks_第11张图片

 maxpooling

CS231n课程笔记:Leture5 Convolutional Neural Networks_第12张图片

 CS231n课程笔记:Leture5 Convolutional Neural Networks_第13张图片

 

你可能感兴趣的:(java,前端,服务器)