computer_vision_learning

Pose Estimation Model: https://google.github.io/mediapipe/solutions/pose.html

autoencoder-decoder to do binary classification for stable and non-stable event.

segment-anything

TODO

• GUI
  • Add menu helper
  • Add current label indicator
  • Add Auto-Refresh
  • Add playback speed controls

Usage

• make test to test different configurations specify in src/test.py

• make record to record video to video/ folder

  • change MAX_TIME, FPS for record time or frame per second in src/convert.py

• make label to show all videos in video/ folder (max=10) with GUI, click 'label' then 'convert' to label each video.

  • something like this
  • 

• make model <file_name_in_nn_folder>

  • make model LSTM to run LSTM.py and save model to model/
  • specify <name>.mp4.csv files in LSTM.py to train

• make predict to test model

  • specify MODEL_NAME to load in src/predict.py
  • might need to retrain model on different machine to get the working keras_metadata.pb file

Visitor Count

Setup

• Windows (recommended)

  • use pip install
    • libraries:
      • opencv-python
      • mediapipe
      • pandas
      • pygame
      • and other modules defined in pyproject.toml
  • download make utils
    • https://sourceforge.net/projects/gnuwin32/files/make/3.81/make-3.81.exe/download?use_mirror=gigenet&download=
  • run make setup in terminal

• MacOS (intel chip only)

  • use pip install
    • libraries:
      • opencv-python
      • mediapipe
      • pandas
      • pygame
      • and other modules defined in pyproject.toml
    • create video/ and data/ folders to store the outputs
    • run commands in Makefile

• RaspberryPi 64bit Lite OS

  • run the following commands

     sudo apt install git
     git clone https://github.com/huangruoqi/computer_vision_learning.git
     cd computer_vision_learning
     mv Makefile4raspi Makefile
     make setup
  

  or install dependencies manually

     sudo apt update
     # build dependencies
     sudo apt-get install -y build-essential tk-dev libncurses5-dev libncursesw5-dev \
     libreadline6-dev libdb5.3-dev libgdbm-dev libsqlite3-dev libssl-dev libbz2-dev \
     libexpat1-dev liblzma-dev zlib1g-dev libffi-dev
  
     # Python:3.7.0
     wget https://www.python.org/ftp/python/3.7.0/Python-3.7.0.tgz
     sudo tar zxf Python-3.7.0.tgz
     cd Python-3.7.0
     sudo ./configure
     sudo make -j 4
     sudo make altinstall
     echo "alias python='/usr/local/bin/python3.7'" >> ~/.bashrc
     source ~/.bashrc
  
     # pip with Python3.7
     wget https://bootstrap.pypa.io/get-pip.py
     python get-pip.py
  
     # opencv, ffmpeg, mediapipe
     sudo apt install -y ffmpeg python3-opencv fswebcam
     wget https://files.seeedstudio.com/ml/mediapipe/mediapipe-0.8-cp37-cp37m-linux_aarch64.whl
     pip3.7 install mediapipe-0.8-cp37-cp37m-linux_aarch64.whl
  
     # [fix] TypeError: Descriptors cannot be created directly
     pip3.7 install protobuf==3.20.*
  
     # other libs
     pip3.7 install pandas pygame
  

  and change URL_PREFIX in ~/.local/lib/python3.7/site-packages/mediapipe/python/solutions/download_utils.py to 'https://storage.googleapis.com/mediapipe-assets/'