Paras ML Models

Overview

Paras (Parking Automation and Reservation Analytics System) uses machine learning models to optimize parking management through real-time detection, prediction, and analytics. This repository documents the ML models integrated into the system.

Features

• Vehicle Detection: Identifies and classifies vehicles using computer vision.
• Parking Slot Availability Prediction: Predicts available slots using time-series data and sensor inputs.
• License Plate Recognition (LPR): Extracts license plate numbers using OCR.
• Anomaly Detection: Detects illegal parking, unauthorized access, and unusual activities.
• Dynamic Pricing Optimization: Recommends pricing based on demand and occupancy trends.

Model Details

1. Vehicle Detection Model

• Algorithm: YOLOv8 (You Only Look Once)
• Training Data: COCO dataset + custom annotated parking images
• Input: CCTV camera feed
• Output: Bounding boxes and vehicle classifications
• Dependencies:
  • ultralytics
  • opencv-python
  • torch

2. Parking Slot Availability Prediction

• Algorithm: LSTM (Long Short-Term Memory)
• Training Data: Historical occupancy data, weather conditions, and traffic patterns
• Input: Time-series occupancy data
• Output: Probability of slot availability at a future timestamp
• Dependencies:
  • tensorflow
  • pandas
  • numpy

3. License Plate Recognition (LPR)

• Algorithm: CRNN (Convolutional Recurrent Neural Network) + Tesseract OCR
• Training Data: OpenLPR dataset
• Input: Cropped vehicle license plate image
• Output: Extracted text (license plate number)
• Dependencies:
  • pytesseract
  • opencv-python
  • torch

4. Anomaly Detection Model

• Algorithm: Autoencoder + Isolation Forest
• Training Data: Normal vs. abnormal parking behaviors
• Input: Surveillance footage and parking history logs
• Output: Anomaly scores and detected suspicious activity
• Dependencies:
  • scikit-learn
  • tensorflow
  • numpy

5. Dynamic Pricing Optimization

• Algorithm: Reinforcement Learning (Deep Q-Networks)
• Training Data: Historical pricing trends, occupancy rates, and external factors (e.g., weather, events)
• Input: Current occupancy and demand levels
• Output: Recommended pricing strategy
• Dependencies:
  • gym
  • stable-baselines3
  • pandas

Installation

To set up the ML models, clone the repository and install the required dependencies:

# Clone the repository
git clone https://github.com/your-repo/paras-ml.git
cd paras-ml

# Install dependencies
pip install -r requirements.txt

Usage

Each model is structured as a separate module. To run a specific model, use the following:

python vehicle_detection.py

For real-time deployment, integrate the models into the main Paras system using APIs or microservices.

Live Demo

Check out the live demo here: Live Demo

Contributors

• Prasanna Pal (CTO) - ML development and optimization
• Chinmay Bhat (CIO) - Hardware integration and sensor data handling

License

This project is licensed under the MIT License.

---