Opencv-healthcare

Algorithms to digitize Healthcare tests (Chair Sit-Stand, Gait Speed Walk Test, Timed Up and Go) for frailty

Tech Stack:

CV Functionality:

• OpenCV
• MediaPipe for Pose estimation

Virtual Environment:

• Running Python3.11

File Directory

external/: Contains external dependencies and submodules.

• pybind11/: Submodule for pybind11, a lightweight header-only library that exposes C++ types in Python and vice versa.

python_arm_curler/: Contains specific algorithms or modules related to the project, possibly for arm curl exercises.

sit_stand_algorithm/: Contains specific algorithms or modules related to sit-stand exercises.

src/: Contains C++ files that is used to call the CV algorithms written in Python. This process is no longer in use.

CMakeLists.txt: The CMake configuration file used for generating build files and managing the build process.

requirements.txt: List of python dependencies used and versions

Run Locally

Clone repository

    git clone https://github.com/brennanleez-coder/opencv-healthcare.git

    cd opencv-healthcare

Setup virtual environment

Name of virtual environment should be myenv

    python3 -m venv myenv

Activate virtual environment

    source myenv/bin/activate

Install python requirements

This command downloads all required python3 libraries including opencv and mediapipe

    pip install -r requirements.txt

Install git submodules (Pybind11) (Optional, if calling from C++ environment). This is no longer in use.

Pybind11 is used to call python files from a C++ environment

    git submodule update --init --recursive

Ensure external/pybind11 folder is created with contents

Generate executable with Cmake (Optional, if calling from C++ environment). This is no longer in use.

From the root directory

    mkdir -p build && cd build

    cmake ..

    make

Download Test videos

Test Videos Place them into the test/ directory

Development Process

1. Develop and Test Algorithms in Jupyter Notebook

-Begin by developing and testing the computer vision (CV) algorithms in Jupyter Notebook. This allows for an interactive environment where you can quickly iterate and visualize the results.

2. Convert to Cython

• Cython is a superset of Python that allows for the writing of C extensions for Python. It is used to speed up Python code by converting it to C code.
• To convert Python code to Cython, create a .pyx file and write the code in Cython syntax. Then, create a setup.py file to compile the Cython code into a shared library.
• To compile the Cython code and move the shared library to the Python path, run the following command:

    ./Cython_algorithms/compile_and_move.sh

Check that there are no errors, warnings are fine. A build_output.log file will be generated in the Cython_algorithms directory. This file contains the output of the compilation process and can be used to debug any issues that arise during the compilation process.

Deploy Algorithms in FastAPI

server/ contains the fastapi app

use the following command to run the server locally:

    cd server
    fastapi dev

Build fastapi app as docker image

    docker build -t image-name -f Dockerfile .

Run the docker image

    docker run -d --name container-name -p 80:80 image-name

Compiling for Linux

To compile for linux, I created another docker container just to compile and then copy the .so file locally. Then move it to my fastapi app. If not done properly, the .so file will not work on the fastapi app that is running inside the docker container.

    docker build cython-app . 
    docker run -d cython-app

Check if the file is inside

    docker exec -it <CONTAINER_ID> /bin/bash

Enter the container

    docker cp CONTAINER_ID:app/File .          

Once the sit_stand_overall.cpython-310-x86_64-linux-gnu.so is created, move it to the fastapi app