👁️ RiskVision — Real-Time AI Violence Detection System

Protecting communities with intelligent, automated threat detection and instant emergency alerts.

RiskVision (internally named EyeView) is a full-stack AI-powered public safety platform that performs real-time violence detection from live camera feeds using YOLO deep learning models. When a threat is detected, it automatically records video evidence, stores it in Firebase Cloud Storage, and dispatches instant SMS and voice call alerts to authorities via Twilio — all visualised on a live React dashboard.

✨ Features

Feature	Description	Status
🔍 Real-Time Detection	YOLO-powered violence detection with low-latency video processing	✅ Active
📹 Automatic Evidence Recording	Incident clips recorded and uploaded to Firebase Cloud Storage instantly	✅ Active
📲 Instant SMS & Voice Alerts	Twilio sends emergency notifications to configured authority numbers	✅ Active
📊 Live Monitoring Dashboard	Real-time React dashboard showing live feed, alerts, and incident history	✅ Active
📱 Mobile Camera Support	Compatible with mobile device camera streams	✅ Active
🔐 Firebase Authentication	Secure access control for the monitoring dashboard	✅ Active
🏙️ Smart City Ready	Designed for deployment in public CCTV and surveillance environments	✅ Active
🎞️ Incident History	Full log of past detections with timestamps, clips, and alert records	✅ Active

🛠️ How It Works

📷 Live Camera Feed
        │
        ▼
  OpenCV captures frames continuously
        │
        ▼
  YOLO model analyses each frame
  for violent behaviour patterns
        │
    ┌───┴───────────────────┐
    │ Violence NOT detected  │  Violence DETECTED
    │ → Continue monitoring  │        │
    └───────────────────────┘        ▼
                              📹 Flask records incident clip
                                      │
                                      ▼
                              ☁️ Clip uploaded to Firebase
                              Cloud Storage with timestamp
                                      │
                                      ▼
                              📲 Twilio fires SMS + Voice Call
                              to emergency contact numbers
                                      │
                                      ▼
                              📊 React dashboard updates
                              with live alert and clip link

🗂️ Project Structure

RiskVision/
│
├── Backend/                        # Python — Flask API + YOLO detection engine
│   ├── eye-view.py                 # Main application entry point
│   ├── requirements.txt            # Python dependencies
│   ├── config/                     # Firebase & Twilio configuration helpers
│   ├── models/                     # YOLO model weights (.pt files)
│   └── eyeview-v2-firebase-adminsdk-*.json   # Firebase Admin SDK credentials
│
├── EyeView-frontend/               # React 19 + Vite 6 — live monitoring dashboard
│   ├── public/                     # Static assets and icons
│   └── src/
│       ├── components/             # Reusable UI — Alert cards, Video feed, Navbar
│       ├── pages/                  # Dashboard, Incident History, Settings, Login
│       ├── styles/                 # Global and component-level styles
│       └── utils/                  # Firebase client SDK, API helpers, formatters
│
├── Contributing.md                 # Contribution guidelines
├── Evaluator_QnA_Script.md         # Q&A script for project evaluators
├── Presentation_Guide.md           # Presentation walkthrough guide
├── project.md                      # Detailed project documentation
├── package.json                    # Root-level scripts & frontend dependencies
├── .gitignore
└── README.md

🧠 Tech Stack

AI & Computer Vision (Backend)

Python 3.8+ — core backend language
Flask — lightweight API server streaming detection results
YOLO (Ultralytics) — real-time object & violence detection model
OpenCV — video frame capture, preprocessing, and clip recording

Frontend Dashboard

React 19 — latest stable React with concurrent features
Vite 6 — blazing-fast dev server and build tool
Tailwind CSS 4 — utility-first responsive styling
Framer Motion — smooth animations and transitions
Axios — HTTP client for backend API calls
React Router DOM 7 — client-side routing
React Hot Toast — real-time alert toast notifications
React Icons — comprehensive icon library

Cloud & Services

Firebase 11 — Cloud Storage for evidence clips + Authentication
Twilio — SMS and programmable voice call emergency alerts
Concurrently — run frontend and backend simultaneously with a single command

🤖 YOLO Detection Engine

The core intelligence of RiskVision is a YOLO (You Only Look Once) deep learning model trained to identify violent human behaviour patterns in real time.

How the Detection Works

Frame Capture — OpenCV reads continuous frames from the configured camera source (webcam, IP camera, or mobile stream)
Inference — Each frame is passed through the YOLO model which outputs bounding boxes and confidence scores for detected classes
Threshold Filtering — Only detections above a configurable confidence threshold trigger the alert pipeline (avoids false positives)
Clip Recording — OpenCV buffers the frames around the detection moment and saves them as a video clip
Cooldown Period — A configurable cooldown prevents repeated alerts for the same ongoing incident

Model Configuration

The YOLO model weights (.pt file) are stored in Backend/models/. You can swap in any custom-trained YOLO model — for example a model fine-tuned on violence-specific datasets — by updating the model path in eye-view.py.

💡 Using a Custom Model: Train your model using Ultralytics YOLO on a violence detection dataset, export the .pt weights, and drop it into Backend/models/.

🔥 Firebase Integration

RiskVision uses Firebase for two critical functions: Cloud Storage for evidence clips and Authentication for dashboard access control.

Firebase Cloud Storage

When violence is detected, the recorded clip is automatically uploaded to Firebase Cloud Storage with a timestamped filename. The React dashboard then fetches and displays these clips in the incident history log.

Firebase Authentication

The monitoring dashboard is protected behind Firebase Authentication so only authorised operators can access live feeds and incident records.

Setting Up Firebase

Go to https://console.firebase.google.com/ and create a new project
Enable Cloud Storage and Authentication (Email/Password or Google Sign-In) in the console
Go to Project Settings → Service Accounts → Generate New Private Key — download the JSON file
Rename the downloaded file and place it in the Backend/ directory
In Project Settings → General → Your Apps, register a Web App and copy the Firebase config object
Add both sets of credentials to your .env files (see below)

⚠️ The Firebase Admin SDK JSON file contains sensitive credentials. Never commit it to Git. Add it to .gitignore.

📲 Twilio Alert System

When RiskVision detects a violent incident, Twilio immediately contacts configured emergency numbers through both SMS and automated voice calls — ensuring rapid response even when no one is watching the dashboard.

Alert Behaviour

Trigger	Alert Type	Recipient
Violence detected (first frame above threshold)	SMS with timestamp & location label	Admin / Authority number
Violence confirmed (clip recorded)	Voice call with automated message	Admin / Authority number
Detection cooldown expired + new incident	Both SMS and Voice Call repeated	Admin / Authority number

Setting Up Twilio

Create a free account at https://www.twilio.com/
In the Twilio Console, note your Account SID and Auth Token
Get a Twilio Phone Number (free trial number works for testing)
Add all three values plus your target alert number to Backend/.env

💡 On a Twilio free trial account, you can only send alerts to verified phone numbers. Upgrade to a paid account for unrestricted alerting in production.

⚙️ Installation

Prerequisites

Python 3.8 or higher
Node.js 18 or higher
npm 9 or higher
A webcam or IP camera feed
A Firebase project (Storage + Authentication enabled)
A Twilio account

Step 1 — Clone the Repository

git clone https://github.com/Lucky-939/RiskVision.git
cd RiskVision

Step 2 — Backend Setup

cd Backend

# Create a Python virtual environment
python -m venv venv

# Activate — Linux / macOS
source venv/bin/activate

# Activate — Windows (PowerShell)
.\venv\Scripts\activate

# Install all Python dependencies
pip install -r requirements.txt

Step 3 — Place Firebase Admin SDK Credentials

Download your Firebase Admin SDK JSON key from the Firebase Console and place it in the Backend/ directory. Update the filename reference in eye-view.py if needed.

Step 4 — Frontend Setup

cd ../EyeView-frontend
npm install

Step 5 — Configure Environment Variables

Create .env files for both backend and frontend (see Environment Configuration below).

🔑 Environment Configuration

Backend — `Backend/.env`

# ── Twilio Emergency Alerts ───────────────────────────
TWILIO_ACCOUNT_SID=your_twilio_account_sid
TWILIO_AUTH_TOKEN=your_twilio_auth_token
TWILIO_PHONE_NUMBER=+1XXXXXXXXXX        # Your Twilio number
ADMIN_PHONE_NUMBER=+91XXXXXXXXXX        # Alert recipient number

# ── Firebase Admin SDK ────────────────────────────────
FIREBASE_CREDENTIALS_PATH=./eyeview-v2-firebase-adminsdk-*.json
FIREBASE_STORAGE_BUCKET=your-project-id.appspot.com

# ── Detection Settings ────────────────────────────────
DETECTION_CONFIDENCE=0.65               # Confidence threshold (0.0 - 1.0)
ALERT_COOLDOWN_SECONDS=30               # Seconds between repeated alerts
CAMERA_SOURCE=0                         # 0 = default webcam, or IP camera URL

Frontend — `EyeView-frontend/.env`

# ── Firebase Web SDK ──────────────────────────────────
VITE_FIREBASE_API_KEY=your_firebase_api_key
VITE_FIREBASE_AUTH_DOMAIN=your-project-id.firebaseapp.com
VITE_FIREBASE_PROJECT_ID=your-project-id
VITE_FIREBASE_STORAGE_BUCKET=your-project-id.appspot.com
VITE_FIREBASE_MESSAGING_SENDER_ID=your_messaging_sender_id
VITE_FIREBASE_APP_ID=your_firebase_app_id

# ── Backend API ───────────────────────────────────────
VITE_API_BASE_URL=http://localhost:5000

⚠️ Security Rules:

Never commit .env files or the Firebase Admin SDK JSON to Git

Ensure both are listed in .gitignore

Use Firebase Storage Security Rules to restrict public access to evidence clips

🚀 Running the Application

Option A — Run Everything with One Command (Recommended)

From the project root:

npm run start:all

This uses concurrently to start both the Flask backend and Vite frontend simultaneously.

Option B — Run Separately

Terminal 1 — Backend:

cd Backend
source venv/bin/activate   # or .\venv\Scripts\activate on Windows
python eye-view.py

Backend runs at: http://localhost:5000

Terminal 2 — Frontend:

cd EyeView-frontend
npm run dev

Dashboard available at: http://localhost:5173

🐛 Troubleshooting

Firebase credentials file not found

Ensure the Admin SDK JSON file is in the Backend/ directory and the path in eye-view.py matches the filename exactly. Run the backend from inside the Backend/ folder so relative paths resolve correctly.

Backend won't start

Check that your virtual environment is activated, all packages are installed via pip install -r requirements.txt, and no other process is using port 5000. Verify Python version is 3.8 or higher with python --version.

Camera not detected

Set CAMERA_SOURCE=0 in your .env for the default system webcam. For an IP camera, provide the RTSP stream URL (e.g., rtsp://192.168.1.1:554/stream). Ensure no other application is locking the camera.

Frontend can't connect to backend

Verify the backend is running on port 5000 and that VITE_API_BASE_URL=http://localhost:5000 is set in EyeView-frontend/.env. Check that CORS is configured in eye-view.py to allow requests from http://localhost:5173.

Twilio alerts not sending on free trial

Free trial accounts can only send to verified numbers. Verify the recipient number in the Twilio Console under Verified Caller IDs, or upgrade to a paid account.

🛣️ Roadmap

Multi-camera support — monitor multiple feeds simultaneously from one dashboard
Advanced AI models — integrate more sophisticated pose estimation for improved accuracy
Analytics dashboard — incident heatmaps, frequency trends, and response time reporting
Mobile app — React Native app for field operators to receive and review alerts
Email notifications — alongside Twilio SMS and voice alerts
Enhanced security — role-based access control with multiple operator levels
Kubernetes deployment — horizontal scaling for enterprise and smart city deployment
Edge computing support — run YOLO inference on-device for reduced latency

🤝 Contributing

Contributions are welcome — especially improvements to detection accuracy, UI enhancements, and performance optimisation.

# 1. Fork the repository on GitHub

# 2. Clone your fork
git clone https://github.com/YOUR-USERNAME/RiskVision.git

# 3. Create a feature branch
git checkout -b feature/your-feature-name

# 4. Make your changes and commit
git add .
git commit -m "feat: describe your change here"

# 5. Push and open a Pull Request
git push origin feature/your-feature-name

Guidelines

Follow PEP 8 for Python code and ESLint rules for JavaScript/React
Use conventional commit messages (feat:, fix:, docs:, perf:)
Comment all YOLO model configuration and detection threshold logic clearly
Test on both webcam and IP camera sources before submitting a PR
For major changes, open an issue first to discuss your approach
See Contributing.md for the full contribution guide

🐛 Reporting Issues

Open an issue with:

A clear title and description of the problem
Steps to reproduce it
Your Python and Node.js versions
Relevant error logs or screenshots

📄 License

This project is licensed under the GNU General Public License v3.0 — see the LICENSE file for details.

Made with ❤️ for a safer world — Lucky-939

Name		Name	Last commit message	Last commit date
Latest commit History 50 Commits
Backend		Backend
EyeView-frontend		EyeView-frontend
node_modules		node_modules
.gitignore		.gitignore
Contributing.md		Contributing.md
Evaluator_QnA_Script.md		Evaluator_QnA_Script.md
LICENSE		LICENSE
Presentation_Guide.md		Presentation_Guide.md
README.md		README.md
package-lock.json		package-lock.json
package.json		package.json
project.md		project.md

Folders and files

Latest commit

History

Repository files navigation

👁️ RiskVision — Real-Time AI Violence Detection System

📌 Table of Contents

✨ Features

🛠️ How It Works

🗂️ Project Structure

🧠 Tech Stack

🤖 YOLO Detection Engine

How the Detection Works

Model Configuration

🔥 Firebase Integration

Firebase Cloud Storage

Firebase Authentication

Setting Up Firebase

📲 Twilio Alert System

Alert Behaviour

Setting Up Twilio

⚙️ Installation

Prerequisites

Step 1 — Clone the Repository

Step 2 — Backend Setup

Step 3 — Place Firebase Admin SDK Credentials

Step 4 — Frontend Setup

Step 5 — Configure Environment Variables

🔑 Environment Configuration

Backend — Backend/.env

Frontend — EyeView-frontend/.env

🚀 Running the Application

Option A — Run Everything with One Command (Recommended)

Option B — Run Separately

🐛 Troubleshooting

🛣️ Roadmap

🤝 Contributing

Guidelines

🐛 Reporting Issues

📄 License

About

Resources

License

Contributing

Uh oh!

Stars

Watchers

Forks

Releases

Packages 0

Uh oh!

Contributors

Uh oh!

Languages

Backend — `Backend/.env`

Frontend — `EyeView-frontend/.env`

Packages