Skip to content

santaboi/AI_CUP_STAS

 
 

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

27 Commits
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

AI CUP STAS

Preprocessing(cross validation)

  • 將資料分群,不同特性的資料最好均勻分布各個cross validation 的資料集 , 以增加ensemble後的模型穩定度
    • way1 : 肉眼觀察分群
    • way2 : kmeans cluster
      • step1 : Elbow Method 決定分4群最佳 (註: 用keras inceptionV3 抽特徵)
      • step2 : 4群均分到不同training dataset Untitled (2) Untitled (3)

image

Model Architecture (cross validation + voting)

image image image image

Postprocessing (blur + findcontours&fillpoly)

image

比賽結果

  • 最後正式公布結果有些組別被移除資格,名次有異動以官網為準 image
  • 官網連結 : AICUP

repo解說

前處理程式碼:Preprocessing.ipynb

訓練程式碼:Model.ipynb

辨識程式碼:Test.ipynb

模型檔案:https://drive.google.com/drive/folders/1Sq682KheFmneDXpLD5Y7cYXpxpHj-gup?usp=sharing

執行環境:TWCC

預測結果輸出:STAS.zip

執行環境:Pytorch 1.11.0

檔案目錄結構

├── Preprocessing.ipynb               
├── Model.ipynb
├── Test.ipynb
├── Data
│   ├── SEG_Train_Datasets
│   │   ├── Train_Images
│   │   ├── Train_Annotations
│   │   ├── Train_Masks
│   │   ├── Fold1_Images
│   │   ├── Fold1_Masks
│   │   ├── Fold2_Images
│   │   ├── Fold2_Masks
│   │   ├── Fold3_Images
│   │   ├── Fold3_Masks
│   │   ├── Fold4_Images
│   │   ├── Fold4_Masks
│   │   ├── Test_Images
│   │   └──  Test_Masks
│   ├── Public_Image
│   └── Image   
├── model_weight                      
│   ├── best_model_1.pth
│   ├── best_model_2.pth               
│   ├── best_model_3.pth         
│   ├── best_model_4.pth               
│   └── best_model_5.pth
├── fig                      
│   ├── acc_1.png
│   ├── acc_2.png              
│   ├── acc_3.png         
│   ├── acc_4.png              
│   ├── acc_5.png
│   ├── loss_1.png
│   ├── loss_2.png              
│   ├── loss_3.png         
│   ├── loss_4.png              
│   └── loss_5.png
├── output                      
│   ├── Image           
│   └── Image_Postprocess

About

No description, website, or topics provided.

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published

Languages

  • Jupyter Notebook 100.0%