Labfans是一个针对大学生、工程师和科研工作者的技术社区。 论坛首页 | 联系我们(Contact Us)
MATLAB爱好者论坛-LabFans.com
返回   MATLAB爱好者论坛-LabFans.com > 其它 > 资料存档
资料存档 资料存档
回复
 
主题工具 显示模式
旧 2019-11-24, 02:25 AM   #1
高级会员
 
注册日期: 2019-11-21
帖子: 2,897
感谢他人: 0
有 0 帖获得 0 感谢
声望力: 60
poster 正向着好的方向发展
默认 Poor results with regression network

I use NASNet (amongst other networks) as regression network. The output consists of 4 real numbers. The input is an 1536x1536 image resized to 331x331.



The output values describe polynomial coefficients. The first two values are much smaller than the last two values (1e-06 to 1e-02).



I've implemented the network in Matlab and used for the first try 125 training and 53 test images.



As a result, the network output values differ extremely from the desired values. All 4 values are in the region of 1e-01 and 1e00.



What can I do to get better results? More training data? Avoiding image resizing? Changing the network architecture? I've only replaced the classification and fully connected layer by a new fully connected layer (with the proper output number) and a regression layer.





More...
poster 当前离线   回复时引用此帖
回复

主题工具
显示模式

发帖规则
不可以发表新主题
不可以发表回复
不可以上传附件
不可以编辑自己的帖子

启用 BB 代码
论坛禁用 表情符号
论坛启用 [IMG] 代码
论坛启用 HTML 代码
Trackbacks are 禁用
Pingbacks are 禁用
Refbacks are 禁用



所有时间均为北京时间。现在的时间是 06:23 AM


Powered by vBulletin
版权所有 ©2000 - 2021,Jelsoft Enterprises Ltd.

SEO by vBSEO ©2009, Crawlability, Inc.