Watchtower Live 🔹

A real-time AI-powered surveillance system that uses your webcam to detect objects and automatically capture images when specific objects are detected.

Features

• Real-time Object Detection: Uses TensorFlow.js and COCO-SSD model to detect objects in real-time
• Automatic Image Capture: Saves images when detected objects meet confidence thresholds
• Customizable Settings: Adjust confidence thresholds, save intervals, and notification intervals
• Push Notifications: Get notified when objects are detected (requires VAPID keys)
• WebSocket Support: Real-time communication between client and server
• Image Gallery: View and manage captured images
• Responsive Design: Works on desktop and mobile devices

Recent Fixes

The following issues have been resolved:

1. JavaScript Errors: Fixed null reference errors by adding proper null checks for DOM elements
2. Missing Settings Panel: Added the complete settings panel HTML that was referenced in JavaScript
3. Camera Access: Improved error handling for camera access issues
4. Favicon 404: Added inline SVG favicon to prevent 404 errors
5. Initialization: Enhanced initialization process with better error handling and logging

Installation

For Desktop/Development:

1. Clone or download this repository

2. Install dependencies:

   npm install

3. (Optional) Set up VAPID keys for push notifications:

   # Generate VAPID keys using web-push
   npx web-push generate-vapid-keys

   Then create a .env file with:

   VAPID_PUBLIC_KEY=your_public_key_here
   VAPID_PRIVATE_KEY=your_private_key_here
   VAPID_EMAIL=your_email@example.com
   

For Raspberry Pi:

See PI_SETUP.md for detailed Raspberry Pi setup instructions.

Quick Pi Setup:

# Install on Pi
git clone <repository-url>
cd watchtower
npm install

# Run with Pi optimizations
./pi-start.sh

Usage

1. Start the server:

   node server.js

2. Open your browser and navigate to http://localhost:3000

3. Allow camera access when prompted

4. Configure your settings:

   • Confidence Threshold: Minimum confidence level for object detection (10-90%)
   • Save Interval: How often to save images (1-10 seconds)
   • Notification Interval: How often to send notifications (10-60 seconds)
   • Objects to Detect: Select which objects to monitor (person, car, dog, cat, cell phone, laptop, bicycle, motorcycle, truck, bus)

5. Use the controls:

   • Toggle Detection Boxes: Show/hide bounding boxes around detected objects
   • Toggle Recording: Enable/disable automatic image capture
   • Test Notification: Send a test push notification
   • Clear All Images: Delete all captured images

How It Works

1. Camera Setup: The application requests access to your webcam
2. AI Model Loading: TensorFlow.js loads the COCO-SSD object detection model
3. Real-time Detection: The model analyzes video frames to detect objects
4. Image Capture: When detected objects meet your criteria, images are automatically saved
5. Notifications: Push notifications are sent when objects are detected
6. WebSocket Communication: Real-time updates are sent between client and server

File Structure

watchtower/
├── server.js              # Main server file
├── package.json           # Dependencies
├── public/
│   ├── index.html         # Main application page
│   ├── sw.js             # Service worker for notifications
│   └── recordings/       # Directory for captured images
└── README.md             # This file

Troubleshooting

Camera Not Working

• Ensure your browser supports getUserMedia
• Check that camera permissions are granted
• Try refreshing the page

No Video Preview

• Check browser console for errors
• Ensure camera is not being used by another application
• Try a different browser

Push Notifications Not Working

• Ensure VAPID keys are properly configured
• Check that notifications are enabled in browser settings
• Verify service worker is registered

JavaScript Errors

• Clear browser cache and refresh
• Check browser console for specific error messages
• Ensure all dependencies are properly loaded

Browser Compatibility

• Chrome/Chromium (recommended)
• Firefox
• Safari (limited support)
• Edge

Security Notes

• This application runs locally and does not send video data to external servers
• Images are stored locally in the public/recordings/ directory
• Push notifications require VAPID keys for security
• Camera access is required for functionality

License

This project is open source and available under the MIT License.