GUARDIAN-EYE.AI

GUARDIAN EYE.AI provides an innovative solution for public safety in urban India, utilizing AI and computer vision to detect violence in real-time.

Comment

Guardian-Eye.AI

Overview

Guardian-Eye.AI is an AI-powered surveillance system designed to enhance public safety by detecting violent activities in real-time. Using computer vision and deep learning, this system integrates with existing surveillance infrastructure to identify threats, send alerts, and enable rapid response to critical incidents.

Demo Video

πŸ“Ί Watch Guardian-Eye.AI in action:

Key Features

  • βœ… Real-Time Violence Detection – Utilizes YOLO (You Only Look Once) for high-speed and accurate detection.
  • βœ… FastAPI-Powered Backend – Provides a lightweight, efficient API for both web and mobile integration.
  • βœ… Live Video Streaming – Streams processed video with detected threats overlaid on the feed.
  • βœ… Incident Logging – Automatically records date, time, severity, and detection confidence for review.
  • βœ… Automated Alerts – Sends real-time alerts via Telegram and triggers emergency calls using Millis AI/Twilio.
  • βœ… Configurable Settings – Adjust alert thresholds, detection confidence, and contact details via an intuitive web dashboard.
  • βœ… Intelligent Alert Escalation – Triggers review alerts for MILD cases and automatically escalates to HIGH severity if 5 consecutive MILD detections occur.
  • βœ… Web Dashboard (HTMX + FastAPI) – Provides a responsive UI with live status updates and incident history.

Services Used

  • YOLO (Ultralytics) – For real-time object detection and violence recognition.
  • FastAPI – Backend framework for handling API requests and responses.
  • HTMX – Enables dynamic UI updates without a full page reload.
  • OpenCV – Processes video frames for real-time analysis.
  • Twilio – Sends emergency phone alerts in case of detected violence.
  • Telegram API – Delivers real-time alerts to predefined chat groups.
  • Millis AI – Automates emergency call initiation based on AI analysis.

Installation & Setup

Prerequisites

Ensure you have the following installed:

  • Python 3.8+
  • pip (Python package manager)
  • Virtual environment (recommended)
  • A working webcam (or adjust the code for a different video source)
  • Git LFS – Required for downloading large model files. Make sure Git LFS is installed before cloning the model repository.

1️⃣ Clone the Repository

git clone https://github.com/SHASWATSINGH3101/GUARDIAN-EYE.AI.git
cd GUARDIAN-EYE.AI

2️⃣ Download Pre-trained Models

Models are hosted on Hugging Face. To download them:

git lfs install
git clone https://huggingface.co/SHASWATSINGH3101/GAURDIAN_EYE_AI

After cloning, update the model paths in your app/config.py:

MODEL1_PATH = "GAURDIAN_EYE_AI/model1/best.pt"
MODEL2_PATH = "GAURDIAN_EYE_AI/model2/best.pt"
MODEL3_PATH = "GAURDIAN_EYE_AI/model3/best.pt"

3️⃣ Set Up a Virtual Environment (Recommended)

python -m venv venv
source venv/bin/activate

4️⃣ Install Dependencies

pip install -r requirements.txt

5️⃣ Setting up Telegram Alerts

  1. Create a new bot on Telegram using BotFather.
  2. Get your TELEGRAM_BOT_TOKEN and TELEGRAM_CHAT_ID.
  3. Update the following in your configuration (e.g., in app/config.py): python TELEGRAM_BOT_TOKEN = "your_bot_token_here" TELEGRAM_CHAT_ID = "your_chat_id_here" Use this video to get your bot's TELEGRAM_CHAT_ID:- Chat ID Tutorial

6️⃣ Using Twilio and Millis AI

For full functionality and control, you must create your own accounts and update your configuration with your personal credentials.

⭐Note:-

  • Make sure to get your emergency number verfied through the twillio dashboard.

### Twilio: Sign up for an account at Twilio and obtain your:

  TWILIO_ACCOUNT_SID
  TWILIO_AUTH_TOKEN
  TWILIO_PHONE_NUMBER

### Millis AI: Sign up for an account at Millis AI and obtain your:

  MILLIS_API_KEY
  MILLIS_AGENT_ID 
  FROM_PHONE_NUMBER 
  TO_PHONE_NUMBER

Update your app/config.py with your personal credentials:

TWILIO_ACCOUNT_SID = "your_account_sid_here"
TWILIO_AUTH_TOKEN = "your_auth_token_here"
TWILIO_PHONE_NUMBER = "your_twilio_number_here"

MILLIS_API_KEY = "your_millis_api_key_here"
MILLIS_AGENT_ID = "your_agent_id_here"
FROM_PHONE_NUMBER = "your_twilio_number_here"
TO_PHONE_NUMBER = "your_emergency_number_here"
  • Voice AI Agent Prompt for Millis AI:

To configure the Millis AI voice agent for reporting violent incidents, use the following prompt:

  You are a violent incident reporting AI agent. Follow these steps during calls:

  Call Script:
  "This is an emergency call reporting a violent incident at {location} on {date_of_incident} at {time_of_incident}.
  The severity level is {severity_level}, with {detections} cases detected.
  The detection confidence is {confidence}%."

  {If applicable, include additional information:}

  Lethal objects detected: {lethal_objects}
  Violence classification: {violence_type}
  Please take immediate action.

  Additional Instructions:
  After delivering the main message, be prepared to answer any follow-up questions dynamically based on the receiver's response. Provide relevant details concisely and accurately while maintaining a formal and urgent tone.

  Note:
  - Be polite and professional with the receiver.
  - Ensure that the information is delivered clearly and accurately.
  - Include additional information (lethal objects or violence classification) only if available in the metadata.

Other resources

7️⃣ Run the Application

uvicorn main_fastapi:app --reload

Troubleshooting Tips

  • Git LFS Issues: Ensure Git LFS is installed before cloning the model repository.
  • Webcam Access: If you encounter errors accessing your webcam, verify that your system permissions are correctly set or modify the video source in the code.
  • Port Conflicts: If the server does not start, ensure port 8000 (or the port specified in your configuration) is not in use or adjust the command accordingly.
  • Dependency Conflicts: Use a virtual environment to prevent package conflicts.

License

This project is licensed under the MIT License.

Citing the project

To cite this repository in publications:

@misc{GUARDIAN-EYE.AI,
  author = {SHASWATSINGH3101,Siddharth-sahu-21, Nikhilverma-codes},
  title = {GUARDIAN-EYE.AI},
  year = {2025},
  howpublished = {\url{https://github.com/SHASWATSINGH3101/GUARDIAN-EYE.AI}},
  note = {GitHub repository},
}

Contact

For inquiries or support, connect via:

  • πŸ’¬ Discord: shaswat_singh

Built With

Share this project:

Updates