YoJa Project

Overview

YoJa is a yoga posture monitoring application that utilizes machine learning, Convolutional Neural Networks (CNN), and OpenCV to identify and correct yoga postures in real-time. The application is designed to run on a local server using Ubuntu, leveraging Docker for containerization to ensure a consistent development environment.

Project Structure

The project is organized into four main components:

• Backend: Built using FastAPI, it handles the application logic, including machine learning model interactions and API endpoints for yoga posture monitoring.
• Frontend: Developed using ReactJS, it provides an interactive user interface for users to interact with the yoga posture monitoring system.
• Database: PostgreSQL is used as the database to store user data, posture information, and other relevant details.
• Machine Learning: Includes scripts and models for training, validating, and deploying the yoga posture detection and correction system.

Technologies Used

Backend

• Python
• FastAPI
• OpenCV
• Machine Learning Libraries (e.g., TensorFlow, PyTorch)
• PostgreSQL (for data storage)

Frontend

• ReactJS
• HTML, CSS, JavaScript

Machine Learning

• TensorFlow
• OpenCV
• Scikit-learn
• Jupyter Notebooks

Containerization

• Docker
• Docker Compose

Setup Instructions

Prerequisites

• Docker and Docker Compose installed on your local machine.
• Basic knowledge of Python, FastAPI, ReactJS, and Docker.

Running the Application

1. Clone the Repository

   git clone <repository-url>
   cd YoJa

2. Build and Run the Docker Containers Navigate to the root directory of the project (where the main docker-compose.yml file is located) and run the following command:

   docker-compose up --build
   

3. Access the Application

   • Backend API: Available at http://localhost:8000.
   • Frontend Application: Accessible at http://localhost:3000.
   • Database: PostgreSQL can be accessed locally on the default port 5432 (or as configured in docker-compose.yml).
   • Machine Learning Workspace:
     • Jupyter Notebook (if enabled) will be accessible at http://localhost:8888.
     • Trained models and scripts are located in the machine_learning directory.

Development

• Backend: Modify the files in the backend/app directory. Use FastAPI for API development and integrate the machine learning models as needed.
• Frontend: Make changes in the frontend/src directory. Use ReactJS to enhance the user interface and ensure seamless interaction with the backend.
• Database: Update the database schema or initial data in the database/init.sql file. Use PostgreSQL for managing user and posture data.
• Machine Learning: Modify or add scripts in the machine_learning/scripts directory for training, validation, or preprocessing. Use Jupyter Notebooks in the machine_learning/notebooks directory for exploratory analysis and model development.

Database Initialization

• The database schema and initial data can be set up by executing the SQL commands in database/init.sql.
• To initialize the database, follow these steps:
  1. Ensure the PostgreSQL container is running by starting the Docker containers:

     docker-compose up --build

  2. Access the PostgreSQL container:

     docker exec -it <database-container-name> psql -U <username> -d <database-name>

     Replace <database-container-name>, <username>, and <database-name> with the appropriate values from your docker-compose.yml file.
  3. Run the SQL commands in init.sql to set up the schema and initial data:

     \i /path/to/init.sql

     Replace /path/to/init.sql with the actual path to the init.sql file inside the container.

Best Practices

• Use version control (e.g., Git) to manage changes to the codebase.
• Write unit tests for both backend and frontend components to ensure code quality.
• Document your code and maintain clear comments for better understanding.
• Regularly update dependencies and monitor for security vulnerabilities.

Contributing

Contributions are welcome! Please submit a pull request or open an issue for any suggestions or improvements.

License

This project is licensed under the MIT ADT University License. See the LICENSE file for more details.