程序師世界是廣大編程愛好者互助、分享、學習的平台,程序師世界有你更精彩!
首頁
編程語言
C語言|JAVA編程
Python編程
網頁編程
ASP編程|PHP編程
JSP編程
數據庫知識
MYSQL數據庫|SqlServer數據庫
Oracle數據庫|DB2數據庫
您现在的位置: 程式師世界 >> 編程語言 >  >> 更多編程語言 >> Python

Python based fashionmnist data analysis and processing classification course paper + project source code and data

編輯:Python

Resource download address :https://download.csdn.net/download/sheziqiong/85718886
Resource download address :https://download.csdn.net/download/sheziqiong/85718886
Catalog
kaggle ranking 2
Data analysis and processing 3
1、 Picture analysis 3
2、 Data partitioning 3
Data to enhance 4
Model selection 4
1、3 Simple model of layer convolution 4
2、VGG16 modify 4
3、DPN modify 5
4、resnet 18 modify 5
5、resnet 50 modify 6
6、Google Net 6
Parameter adjustment 6
1、 Learning rate and optimizer 6
2、 Data batch size batch size 7
3、 Rotation Angle 7
4、 Model parameters 7
Training strategy 8
1、 Control random seeds 8
2、 Preliminary training 8
3、Bagging vote 8
summary 9
Reference material 9
kaggle ranking
user name :HyperionSoldier
The user name implies : Science fiction 《 Hei Leian 》 Shrike is the ultimate agent created by human beings in the future , With “ Heberian warriors ” Its name symbolizes the powerful power and use of artificial intelligence AI The determination to solve problems .
private A list of :Score 0.91514 Ranking the first 19

1、3 Simple model of layer convolution
First of all, there is 3 A simple model of layer convolution , Convolution between batchnorm Layer and the relu Layer connection , Improve the generalization ability of the model .
This simple model is used in validation The highest set can reach 90.9% The accuracy of , stay kaggle The public list can reach 87% The accuracy of .
2、VGG16 modify
VGG Models are often used for various classification tasks , Have obtained 2014 year ILSVRC Second place in the competition , Here is VGG Parameter configuration of the model (Simonyan & Zisserman, 2014).
I used... For this task VGG16 Modified version of ( In the table C), First of all, will RGB 3 The input of the channel is changed to 1 Channel grayscale , And finally 3 Convolutions and maxpool Remove the pool layer , Finally, change the dimension of the full connection layer to 256-
512-10






Resource download address :https://download.csdn.net/download/sheziqiong/85718886
Resource download address :https://download.csdn.net/download/sheziqiong/85718886


  1. 上一篇文章:
  2. 下一篇文章:
Copyright © 程式師世界 All Rights Reserved