parkit! 🚗

A backend only (for now) modern social media platform that gamifies parking by rating parking jobs using computer vision and machine learning. Users can post pictures of parked cars, and our system automatically detects vehicles, recognizes license plates, and calculates a parking score based on alignment within parking spaces.

🌟 Features

• Intelligent Parking Analysis

  • Vehicle detection using YOLOv8
  • License plate recognition with FastANPR
  • Automated parking score calculation based on alignment
  • Vehicle color and model recognition

• Social Platform

  • User profiles and authentication via Google OAuth2
  • Photo sharing and scoring system
  • Community engagement through parking ratings
  • User statistics and leaderboards

• Secure(ish) Backend

  • RESTful API built with FastAPI
  • PostgreSQL database for data persistence
  • JWT-based authentication
  • Kubernetes-ready deployment configuration

🔧 Technical Architecture

Backend Stack

• Web Framework: FastAPI
• Database: PostgreSQL
• Authentication: Google OAuth2
• ML/CV Components:
  • YOLOv8 for vehicle detection
  • FastANPR for license plate recognition
  • TensorFlow for vehicle color/model classification
• Deployment: Kubernetes

Directory Structure

backend/
├── app/
│   ├── auth.py           # Authentication logic
│   ├── database.py       # Database models and configuration
│   ├── main.py          # FastAPI application and routes
│   ├── parking_rating.py # Parking score calculation
│   ├── posts.py         # Post handling logic
│   └── utils.py         # Utility functions
├── k8s/                 # Kubernetes configuration files
└── Dockerfile          # Container configuration

🚀 Getting Started

Prerequisites

• Python 3.7+
• PostgreSQL
• Kubernetes cluster (for deployment)
• Google OAuth2 credentials

Local Development Setup

1. Clone the repository:

git clone https://github.com/yourusername/parkit.git
cd parkit

2. Create and activate a virtual environment:

python -m venv venv
source venv/bin/activate  # On Windows: .\venv\Scripts\activate

3. Install dependencies:

pip install -r backend/app/requirements.txt

4. Set up environment variables:

export GOOGLE_CLIENT_ID="your_client_id"
export GOOGLE_CLIENT_SECRET="your_client_secret"
export GOOGLE_REDIRECT_URI="your_redirect_uri"

5. Start the development server:

cd backend/app
uvicorn main:app --reload

Kubernetes Deployment

1. Create the namespace:

kubectl apply -f backend/k8s/ns.yml

2. Deploy PostgreSQL:

kubectl apply -f backend/k8s/postgres-configmap.yml
kubectl apply -f backend/k8s/postgres-deployment.yml
kubectl apply -f backend/k8s/postgres-service.yml

3. Deploy the application:

kubectl apply -f backend/k8s/deployment.yml
kubectl apply -f backend/k8s/parkit-service.yml

The MAKEFILE is specific to OpenRC and minikube.

📈 Development Roadmap

In Progress

• Fix current bugs and tidy app tech debt.
• Improved error handling and input validation
• User profile customization
• Enhanced parking score algorithm
• API documentation with Swagger/OpenAPI

Planned Features

• Swift frontend

🐛 Known Issues

1. Security Concerns

   • JWT secret key configuration needs improvement
   • Sensitive information should be moved to environment variables

2. Technical Debt

   • Need better error handling
   • Lack of comprehensive test coverage
   • vehicle_detection_tracker integration needs refinement

📝 License

This project is licensed under the MIT License - see the LICENSE file for details.

🙏 Acknowledgments

• vehicle_detection_tracker for vehicle detection capabilities
• YOLOv8 team for the object detection model
• FastAPI team for the excellent web framework
• The open source community for various tools and libraries used in this project