Data

Tools

Indexes

Applications

Experiments

Awesome

An open API service indexing awesome lists of open source software.

https://github.com/ziweek/desirable-sea

πŸ† Grand Prize, in 2023 Jeju Island Satellite Data Utilization Competition - πŸ›°οΈπŸ—ΊοΈ Intelligent Platform for Long-Term Abandoned Vehicle Detection Utilizing Jeju Satellite Data
https://github.com/ziweek/desirable-sea

Last synced: 2 months ago
JSON representation

πŸ† Grand Prize, in 2023 Jeju Island Satellite Data Utilization Competition - πŸ›°οΈπŸ—ΊοΈ Intelligent Platform for Long-Term Abandoned Vehicle Detection Utilizing Jeju Satellite Data

Awesome Lists containing this project

README 

        
# Hello, Desirable Sea








  

    ENGLISH

  

  Β |Β 

  

    KOREAN

  

  


  


  πŸ† Grand Prize, in 2023 Jeju Island Satellite Data Utilization Competition πŸ†

  


  


  Intelligent Platform for Long-Term Abandoned Vehicle Detection Utilizing Jeju Satellite Data

  


  



  

    

  

  

    

  


  

    

  

  

    

  

  

    

  

  

    

  

  



  

   

  

  

  

  

  


  

  

  

  

  


  


  

  

  









  

  

  Check out prototypes in the badge below.

  


  


  

    

  

  

    

  

  

    

  











# 1. Introduction



> [!NOTE]

>

> - This idea is a project that was submitted to the [2023 Jeju Satellite Data Utilization Contest](https://aifactory.space/task/2700/overview) hosted by Jeju Special Self-Governing Province and won the Grand Prize (Jeju Special Self-Governing Province Governor Award).

> - This idea is to research and develop an intelligent platform that utilizes Jeju satellite data and applies computer vision object recognition AI technology to detect vehicles parked or abandoned for a long time in the Jeju area and to integrate and manage the information by related organizations.



 Why is it needed?

 There is serious issues caused surrounding parking lots in JeJu island. And also there are heavy traffic problems. 

details>



https://github.com/ziweek/desirable-sea/assets/99459331/4a7edb47-754c-47fd-8dab-7d2adbb045ec



  

     

      


        Image Super Resolution

      


    

    

      


        Small Object Detection

      


    

    

      


        Prototype Web Application

      


    

  

   

    

      

    

    

      

    

    

      

    

  














# 2. Implementation



 핡심기λŠ₯



### 1. Image Super Resolution



 

  

     

      
1.1 Image Super Resolution


    

  

   

     

      

    

  

   

    

     
μ†Œν˜• 객체 탐지 λͺ¨λΈμ˜ μ„±λŠ₯을 κ°œμ„ ν•˜κΈ° μœ„ν•΄ μ „μ²˜λ¦¬ λͺ¨λΈλ‘œ XPixelGroup의 HATλΌλŠ” μ΄ˆκ³ ν•΄μƒλ„ κ°œμ„  λͺ¨λΈμ„ λ„μž…ν•˜μ˜€μŠ΅λ‹ˆλ‹€. 이λ₯Ό 톡해 λŒ€νšŒ μΈ‘μ—μ„œ μ œκ³΅ν•œ μœ„μ„±μ΄λ―Έμ§€ λ°μ΄ν„°μ˜ 해상도λ₯Ό 4λ°° 이상 κ°œμ„ ν•  수 μžˆμ—ˆμŠ΅λ‹ˆλ‹€.


    

  







  

     

      
1.2 μ΄ˆκ³ ν•΄μƒλ„ 이미지 κ°œμ„  λ”₯λŸ¬λ‹ λͺ¨λΈ


    

  

  

   



https://github.com/ziweek/desirable-sea/assets/99459331/2212ce20-bb1b-42f8-a97c-bebd22ce3fb3



    

      

     

      
μ΄ˆκ³ ν•΄μƒλ„ κ°œμ„  λͺ¨λΈλ‘œ μ „μ²˜λ¦¬λœ 데이터을 λΌλ²¨λ§ν•˜μ˜€μŠ΅λ‹ˆλ‹€. μ΄λ•Œ, 사전 ν•™μŠ΅λœ λͺ¨λΈμ„ μ‚¬μš©ν•˜μ—¬ μžλ™μœΌλ‘œ μ°¨λŸ‰μ„ μ‹λ³„ν•˜κ²Œ ν•˜μ˜€μŠ΅λ‹ˆλ‹€. μœ„μ˜ μ˜μƒμ€ 사전 ν•™μŠ΅ λͺ¨λΈμ΄ μ‹λ³„ν•˜μ§€ λͺ»ν•œ μ˜ˆμ™Έμ μΈ κ²½μš°μ— μˆ˜μž‘μ—…μœΌλ‘œ 라벨링을 μ§„ν–‰ν•œ κ²ƒμž…λ‹ˆλ‹€.


    

    










### 2. μ†Œν˜• 객체 탐지 λ”₯λŸ¬λ‹ λͺ¨λΈ



 

  

     

      
2.1 객체 탐지 λ”₯λŸ¬λ‹ λͺ¨λΈ λ„μž…


    

  

   

     

      

    

  

     

    

      
MMRotate ν”„λ ˆμž„μ›Œν¬λ₯Ό ν™œμš©ν•˜μ—¬ μœ„μ„±μ΄λ―Έμ§€ 데이터 λ‚΄μ—μ„œ μ°¨λŸ‰ λ“±μ˜ 객체λ₯Ό 식별할 수 μžˆμ—ˆμŠ΅λ‹ˆλ‹€. μ²˜μŒμ—λŠ” 사전 ν•™μŠ΅λœ λͺ¨λΈμ„ ν™œμš©ν•˜μ—¬ μ„±λŠ₯μ§€ν‘œλ₯Ό ν‰κ°€ν•˜μ˜€μŠ΅λ‹ˆλ‹€.


    

  






 

  

     

      
2.2 Weighted Boxed Ensemble 앙상블 κΈ°λ²•μœΌλ‘œ μ„±λŠ₯ κ°œμ„ 


    

  

    

     

        

    

  

    

    

     

      
μ†Œν˜• 객체 탐지 λͺ¨λΈμ˜ μ„±λŠ₯을 κ°œμ„ ν•˜κΈ° μœ„ν•˜μ—¬, 좔가적인 ν•™μŠ΅κ³Ό Weighted Boxed Ensemble 기법을 μ μš©ν•˜μ˜€μŠ΅λ‹ˆλ‹€. κ·Έ κ²°κ³Ό, 이전에 μ‹λ³„ν•œ μ°¨λŸ‰μ— λŒ€ν•΄μ„œλŠ” 신뒰도가 λŒ€ν­ μƒμŠΉν•œ 것을 확인할 수 μžˆμ—ˆμŠ΅λ‹ˆλ‹€. 뿐만 μ•„λ‹ˆλΌ, μ΄μ „μ—λŠ” μ‹λ³„ν•˜μ§€ λͺ»ν–ˆλ˜ μœ μ±„μƒ‰μ˜ μ°¨λŸ‰μ΄λ‚˜ μ£Όλ³€ ꡬ쑰물의 κ·Έλ¦Όμžκ°€ 결쳐진 μ°¨λŸ‰λ„ 식별할 수 μžˆμ—ˆμŠ΅λ‹ˆλ‹€.


    

  










### 3. μ‚¬μš©μž κ²½ν—˜



 

  

     

      
3.1 Google Map APIλ₯Ό ν™œμš©ν•œ UX κ΅¬ν˜„


    

  

   

     

      

    

  

    

     
Google Map APIλ₯Ό ν™œμš©ν•˜μ—¬ 지도 기반의 μΈν„°νŽ˜μ΄μŠ€λ₯Ό κ΅¬ν˜„ν•˜μ˜€μŠ΅λ‹ˆλ‹€. 이 덕뢄에 μ‚¬μš©μžλŠ” μ œμ£Όλ„ 전역에 μœ„μΉ˜ν•œ 곡영주차μž₯의 μœ„μΉ˜λ₯Ό ν•œλˆˆμ— νŒŒμ•…ν•  수 μžˆμŠ΅λ‹ˆλ‹€. 뿐만 μ•„λ‹ˆλΌ, 좔후에 경둜 탐색 κΈ°λŠ₯을 μΆ”κ°€ν•  κ³„νšμž…λ‹ˆλ‹€.


    

  

  






 

  

     

      
3.2 λ°˜μ‘ν˜• UI 및 닀크λͺ¨λ“œ μ§€μ›μœΌλ‘œ μ‚¬μš©μ„± κ°œμ„ 


    

  

    

     

       

    

  

  

     

      
슀마트폰, νƒœλΈ”λ¦Ώ, λ˜λŠ” μ»΄ν“¨ν„°μ—μ„œλ„ 연속적인 μ‚¬μš©μž κ²½ν—˜μ„ 전달할 수 μžˆλ„λ‘ λ°˜μ‘ν˜• μ›Ήλ””μžμΈμ„ μ μš©ν•˜μ˜€μŠ΅λ‹ˆλ‹€. 또, 더 λ‚˜μ•„κ°€ 닀크λͺ¨λ“œ λ””μžμΈμ„ μ μš©ν•˜μ˜€μŠ΅λ‹ˆλ‹€.


    

  










  μ•„ν‚€ν…μ²˜



### ν”„λ‘œλ•νŠΈ μ•„ν‚€ν…μ²˜



 

   

     

      

    

   

  

    

     
λ³Έ ν”„λ‘œμ νŠΈμ˜ μ„œλΉ„μŠ€ μ•„ν‚€ν…μ²˜λŠ” 크게 4κ°€μ§€λ‘œ λ‚˜λˆ μ„œ μ‚΄νŽ΄λ³Ό 수 μžˆμŠ΅λ‹ˆλ‹€.


     
 1. 데이터 μ „μ²˜λ¦¬ μˆ˜ν–‰ μ„œλ²„


     
    

          
  • μ΄ˆκ³ ν•΄μƒλ„ 이미지 κ°œμ„  λ”₯λŸ¬λ‹ λͺ¨λΈ
    • 

          
  • 핡심 κΈ°λŠ₯ μˆ˜ν–‰ μ„œλ²„(μ†Œν˜• 객체 식별 λ”₯λŸ¬λ‹ λͺ¨λΈ)
    • 

          
  • μ›Ή μ–΄ν”Œλ¦¬μΌ€μ΄μ…˜(ν”„λ‘ νŠΈμ—”λ“œμ™€ λ²‘μ—”λ“œ 및 λ°μ΄ν„°λ² μ΄μŠ€)
    • 

         


     
 2. 핡심 κΈ°λŠ₯ μˆ˜ν–‰ μ„œλ²„


     
    

          
  • μ†Œν˜• 객체 식별 λ”₯λŸ¬λ‹ λͺ¨λΈ
    • 

          
  • 핡심 κΈ°λŠ₯ μˆ˜ν–‰ μ„œλ²„(μ†Œν˜• 객체 식별 λ”₯λŸ¬λ‹ λͺ¨λΈ)
    • 

          
  • μ›Ή μ–΄ν”Œλ¦¬μΌ€μ΄μ…˜(ν”„λ‘ νŠΈμ—”λ“œμ™€ λ²‘μ—”λ“œ 및 λ°μ΄ν„°λ² μ΄μŠ€)
    • 

         


     
 3. μ›Ή μ–΄ν”Œλ¦¬μΌ€μ΄μ…˜


     
    

          
  • ν”„λ‘ νŠΈμ—”λ“œ
    • 

          
  • λ²‘μ—”λ“œ
    • 

          
  • λ°μ΄ν„°λ² μ΄μŠ€
    • 

         


     
 4. μ»¨ν…Œμ΄λ„ˆν™” 및 배포 운영


     
    

          
  • ν”„λ‘ νŠΈμ—”λ“œ
    • 

          
  • λ²‘μ—”λ“œ
    • 

          
  • λ°μ΄ν„°λ² μ΄μŠ€
    • 

         


    

  










### AI μ•„ν‚€ν…μ²˜



 

   

     

      

    

   

  

    

     
λ³Έ ν”„λ‘œμ νŠΈμ˜ AI μ•„ν‚€ν…μ²˜λŠ” 크게 2κ°€μ§€λ‘œ λ‚˜λˆ μ„œ μ‚΄νŽ΄λ³Ό 수 μžˆμŠ΅λ‹ˆλ‹€.


     
 1. μ΄ˆκ³ ν•΄μƒλ„ 이미지 κ°œμ„  λ”₯λŸ¬λ‹ λͺ¨λΈ


     
    

          
  • 

           디렉토리 ꡬ쑰

            

    ```

    ESRGAN

    β”œβ”€β”€ LR                      # Pytorch λͺ¨λΈμ„ μ €μž₯ν•˜λŠ” ν΄λ”μž…λ‹ˆλ‹€.

    β”‚   └── demo.png            # MVP μ‹œμ—°μ— μ‚¬μš©λ  데λͺ¨ μ΄λ―Έμ§€μž…λ‹ˆλ‹€.

    β”œβ”€β”€ models

    β”‚   β”œβ”€β”€ ...

    β”‚   └── RRDB_ESRGAN_x4.pth  # 베이슀라인 κ΄€λ ¨ μ½”λ“œμž…λ‹ˆλ‹€.

    β”œβ”€β”€ ...

    └── results

        └── demo_rlt.png        # λͺ¨λΈλ§ 결과둜 좜λ ₯λ˜λŠ” μ΄λ―Έμ§€μž…λ‹ˆλ‹€.

    ```
    


    • 

  • 

    초기 μ„€μΉ˜λ°©λ²•
    


    ```shell

    # setup.sh 파일 λ§Œλ“€κΈ°.

    # μ•„λž˜μ˜ μ‰˜μŠ€ν¬λ¦½νŠΈ 볡사 ν›„ λΆ™μ—¬λ„£κΈ°.

    # ν„°λ―Έλ„μ—μ„œ setup.sh μ‹€ν–‰
    


    git clone https://github.com/xinntao/ESRGAN

    cd ESRGAN
    


    cd models

    wget --load-cookies ~/cookies.txt \

    "https://docs.google.com/uc?export=download&confirm=$(wget \

    --quiet --save-cookies ~/cookies.txt \

    --keep-session-cookies --no-check-certificate \

    'https://docs.google.com/uc?export=download&id=1TPrz5QKd8DHHt1k8SRtm6tMiPjz_Qene' -O- \

    | sed -rn 's/.*confirm=([0-9A-Za-z_]+).*/\1\n/p')&id=1TPrz5QKd8DHHt1k8SRtm6tMiPjz_Qene" -O RRDB_ESRGAN_x4.pth \

    && rm -rf ~/cookies.txt

    cd ..
    


    cd LR

    wget https://raw.githubusercontent.com/ziweek/desirable-sea/main/src/demo.png

    cd ..

    ```
    


    • 

  • 

    μΆ”λ‘  싀행방법
    


    ```shell

    # 1. μ•„λž˜μ˜ μ‰˜μŠ€ν¬λ¦½νŠΈ 볡사 ν›„ 터미널에 λΆ™μ—¬λ„£κΈ°.

    # 2. ν„°λ―Έλ„μ—μ„œ μ‹€ν–‰ν•˜κΈ°.

    # 3. 동일 디렉토리 λ‚΄μ—μ„œ result-resolution.jpg 파일 ν™•μΈν•˜κΈ°.
    


    python test.py

    ```
    


    • 

         


     
 2. μ†Œν˜• 객체 식별 λ”₯λŸ¬λ‹ λͺ¨λΈ


     
    

          
  • 

    디렉토리 ꡬ쑰

      

    ```

    desirable-sea

    β”œβ”€β”€ checkpoints  # Pytorch λͺ¨λΈμ„ μ €μž₯ν•˜λŠ” ν΄λ”μž…λ‹ˆλ‹€.

    β”œβ”€β”€ mmrotate     # 베이슀라인 κ΄€λ ¨ μ½”λ“œμž…λ‹ˆλ‹€.

    β”‚   β”œβ”€β”€ config

    β”‚   β”œβ”€β”€ ...

    β”‚   └── demo

    β”œβ”€β”€ ...

    β”œβ”€β”€ demo.png     # MVP μ‹œμ—°μ— μ‚¬μš©λ  데λͺ¨ μ΄λ―Έμ§€μž…λ‹ˆλ‹€.

    └── result.jpg   # λͺ¨λΈλ§ 결과둜 좜λ ₯λ˜λŠ” μ΄λ―Έμ§€μž…λ‹ˆλ‹€.

    ```

    • 

  • 

    초기 μ„€μΉ˜λ°©λ²•

     

    ```shell

    # 1. setup.sh 파일 λ§Œλ“€κΈ°.

    # 2. μ•„λž˜μ˜ μ‰˜μŠ€ν¬λ¦½νŠΈ 볡사 ν›„ λΆ™μ—¬λ„£κΈ°.

    # 3. ν„°λ―Έλ„μ—μ„œ setup.sh μ‹€ν–‰
    


    sudo apt update

    sudo apt -y install libgl1-mesa-glx
    


    wget https://repo.anaconda.com/miniconda/Miniconda3-py37_4.9.2-Linux-x86_64.sh

    chmod +x Miniconda3-py37_4.9.2-Linux-x86_64.sh

    bash ./Miniconda3-py37_4.9.2-Linux-x86_64.sh -b -f -p /usr/local

    which conda

    conda --version
    


    pip install torch==1.7.0+cu110 torchvision==0.8.1+cu110 torchaudio==0.7.0 -f https://download.pytorch.org/whl/torch_stable.html

    pip install openmim --use-feature=2020-resolver

    mim install mmcv-full==1.5.3

    mim install mmdet==2.25.1
    


    git clone https://github.com/open-mmlab/mmrotate.git

    cd mmrotate

    pip install -r requirements/build.txt

    pip install -v -e .
    


    cd ..
    


    mkdir checkpoints

    cd checkpoints

    wget https://download.openmmlab.com/mmrotate/v0.1.0/redet/redet_re50_fpn_1x_dota_ms_rr_le90/redet_re50_fpn_1x_dota_ms_rr_le90-fc9217b5.pth

    cd ..
    


    wget https://raw.githubusercontent.com/ziweek/desirable-sea/main/src/demo.png
    


    ````

    • 

  • 

    μΆ”λ‘  싀행방법
    


    ```shell

    # 1. μ•„λž˜μ˜ μ‰˜μŠ€ν¬λ¦½νŠΈ 볡사 ν›„ 터미널에 λΆ™μ—¬λ„£κΈ°.

    # 2. ν„°λ―Έλ„μ—μ„œ μ‹€ν–‰ν•˜κΈ°.

    # 3. 동일 디렉토리 λ‚΄μ—μ„œ result.jpg 파일 ν™•μΈν•˜κΈ°.
    


    python mmrotate/demo/image_demo.py demo.png \

    mmrotate/configs/redet/redet_re50_refpn_1x_dota_ms_rr_le90.py \

    checkpoints/redet_re50_fpn_1x_dota_ms_rr_le90-fc9217b5.pth \

    --out-file result.jpg

    ````
    


    • 

         


    

  










# 3. νŒ€μ›



  

    

      
κΉ€μ§€μš±


    

    

      


        

        

          

        

      


    

    

      


        이 ν”„λ‘œμ νŠΈμ— λ¦¬λ”λ‘œ μ°Έμ—¬ν•˜μ—¬ λ¨Όμ €, μž₯κΈ° 방치 μ°¨λŸ‰ λ¬Έμ œμ— 직면해 이λ₯Ό 관리 ν•˜κΈ° μœ„ν•œ 지λŠ₯ν˜• ν”Œλž«νΌμ˜ ν•„μš”μ„±μ„ κΉ¨λ‹«κ²Œ λ˜μ—ˆμŠ΅λ‹ˆλ‹€. Next.js, Google Map API, 그리고 FastAPI λ“±μ˜ ν˜„λŒ€μ μΈ κΈ°μˆ λ“€μ„ ν†΅ν•©ν•¨μœΌλ‘œμ¨ 우리의 μ†”λ£¨μ…˜μ€ 효율적이고 μ‚¬μš©μž μΉœν™”μ μΈ μž₯κΈ° 방치 μ°¨λŸ‰ 관리λ₯Ό μ‹€ν˜„ν•˜λ©°, μœ„μ„±λ°μ΄ν„°λ₯Ό ν™œμš©ν•œ 지λŠ₯ν˜• ν”Œλž«νΌμ˜ λ°œμ „μ— μΌμ‘°ν•˜λŠ” 데 μžλΆ€μ‹¬μ„ λŠλ‚„ 수 μžˆμ—ˆμŠ΅λ‹ˆλ‹€. 이 ν”Œλž«νΌμ€ 제주의 μ°¨λŸ‰λ¬Έμ œλ₯Ό ν˜μ‹ ν•˜κ³  미래 λ„μ‹œ κ³„νšμ— κΈ°μ—¬ν•˜λŠ” μ€‘μš”ν•œ λ°œνŒμœΌλ‘œμ„œμ˜ 역할을 ν•  κ²ƒμœΌλ‘œ κΈ°λŒ€λ©λ‹ˆλ‹€.

      


    

  

  

   

    

      
λΆ€μ„ μ›…


    

    

      


        

        

          

        

      


    

    

      


        CTO둜 μ°Έμ—¬ν•œ 이 ν”„λ‘œμ νŠΈλŠ” μ΅œμ‹  κΈ°μˆ μ„ ν™œμš©ν•˜μ—¬ 제주의 μž₯κΈ° 방치 μ°¨λŸ‰ λ¬Έμ œμ— λ„μ „ν•˜λŠ” ν₯미둜운 κ²½ν—˜μ΄μ—ˆμŠ΅λ‹ˆλ‹€. μœ„μ„±λ°μ΄ν„°μ™€ λ”₯λŸ¬λ‹μ„ κ²°ν•©ν•˜μ—¬ μ°¨λŸ‰μ„ μ •ν™•ν•˜κ²Œ νƒμ§€ν•˜λŠ” μ‹œμŠ€ν…œμ„ κ°œλ°œν•˜λŠ” κ³Όμ •μ—μ„œ, 기술적인 도전과 ν˜μ‹ μ˜ 기회λ₯Ό λ§ˆμ£Όν•˜κ²Œ λ˜μ—ˆμŠ΅λ‹ˆλ‹€. 이 ν”Œλž«νΌμ€ 미래 제주의 μŠ€λ§ˆνŠΈν•œ 지λŠ₯ν˜• μ‹œμŠ€ν…œμ„ λͺ¨μƒ‰ν•˜λŠ” 데 μžˆμ–΄ 핡심적인 역할을 ν•  κ²ƒμœΌλ‘œ μžλΆ€ν•˜κ³  μžˆμŠ΅λ‹ˆλ‹€.