Skip to content

Two-Step Enhanced Deep Learning Approach for Electromagnetic Inverse Scattering Problems

Notifications You must be signed in to change notification settings

samsam6414/Deep_Learning_Two-Steps

Repository files navigation

Two-Step Enhanced Deep Learning

Two-Step Enhanced Deep Learning Approach for Electromagnetic Inverse Scattering Problems paper:https://ieeexplore.ieee.org/document/8747485

#First Step

image

#Second Step

image

#資料增強法

由於原始的資料量太少,導致在經過2次的訓量後,training 和 testing量大幅下降,利用旋轉矩正的方式來增加:每次選轉90度、旋轉3次以及上下左右翻轉,讓原來的資料成長為12倍

image

image

#Testing Data 增強前

image

#Testing Data 增強後

image

#Testing Data 增強前----------------------------------------------------------------------------#Testing Data 增之後

image 資料在增強後結構相似度(SSIM)明顯提升,誤差值(RMSE)也下降許多。

About

Two-Step Enhanced Deep Learning Approach for Electromagnetic Inverse Scattering Problems

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

 
 
 

Languages