Sentinel

This application uses YOLOv5 for weapon detection and Haar Cascade for facial recognition, capturing webcam feeds to detect threats. It sends real-time email alerts and logs events with timestamps and locations to enhance public safety.

Real-Time Weapon Detection and Facial Recognition System

Overview

This project aims to enhance public safety by developing a real-time weapon detection and facial recognition system. Utilizing YOLOv5 for weapon detection and a Haar Cascade model for facial recognition, this application is designed to monitor webcam feeds, detect weapons and faces, and send real-time alerts. Alerts are also logged with timestamps and locations in a CSV file.

Features

• Real-Time Weapon Detection: Detects weapons in video feeds using YOLOv5.
• Real-Time Facial Recognition: Detects faces using the Haar Cascade model.
• Alert System: Sends email alerts via Gmail and logs alerts with timestamps and locations in a CSV file.
• Webcam Integration: Captures real-time video from the default webcam.

Installation

1. Clone the Repository

   git clone https://github.com/yourusername/weapon-detection-facial-recognition.git
   cd weapon-detection-facial-recognition

2. Set Up Environment

   Create and activate a virtual environment:

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

3. Install Dependencies

   pip install -r requirements.txt

4. Download YOLOv5 and Haar Cascade Models

   • YOLOv5: Clone the YOLOv5 repository into your project directory:

     git clone https://github.com/ultralytics/yolov5.git

   • Haar Cascade: Download the Haar Cascade XML file from the OpenCV repository or use the one provided in the models directory.

5. Configure Email Alerts

   Update the config.py file with your Gmail credentials and other configuration details:

   EMAIL_ADDRESS = 'your-email@gmail.com'
   EMAIL_PASSWORD = 'your-password'

Usage

1. Run the Application

   CSV FILE.py

2. Stop the Application

   The application can be terminated manually or based on elapsed time.

Code Structure

• Exploratory Data Analysis
• Model Training
• Main script to start the real-time detection and alerting system.
• Contains functions for weapon detection and facial recognition.
• Manages the alert system and logging to CSV.
• Configuration file for email alerts and other settings.
• Directory containing the YOLOv5 and Haar Cascade model files.

Challenges and Solutions

• Training Time and Performance: Addressed by fine-tuning with pre-trained weights and using hardware acceleration.
• Alert Delivery: Overcame Gmail authentication issues with proper authentication flows and token management.
• Real-Time Processing: Optimized model inference and video handling for improved performance.

Future Work

• Enhanced Accuracy: Explore advanced models and techniques.
• Location Tracking: Integrate GPS or IP-based location services.
• User Interface: Develop a user interface for easier management and monitoring.
• Social Media Analysis: Implement natural language processing for predictive analysis of unrest.
• Agitator Identification: Integrate face recognition models for identifying known agitators.

Contributing

Contributions are welcome! Please fork the repository and submit a pull request with your changes.

Disclaimer: This project is for educational and research purposes only. Ensure to comply with all local laws and regulations related to surveillance and privacy.