Smart AI Vision

Smart AI Vision is a Progressive Web App (PWA) that performs image recognition using a custom-trained AI model. The application can analyze images from a live camera feed or uploaded files.

Live Demo

You can try out a live version of Smart AI Vision here.

Current AI Model

Custom Model

• Platform: Trained on Teachable Machine by Google
• Model URL: https://teachablemachine.withgoogle.com/models/-fCow7Sym/
• Training Data: 30 images per class (90 total images)
• Classes: 3 trained classes

Features

• Real-time Camera Analysis: Use your device's camera to capture and analyze images instantly
• Image Upload Support: Upload photos from your gallery for analysis
• Progressive Web App:
  • Install directly from your browser
  • Works offline after initial load
  • Native app-like experience on mobile and desktop
• Privacy-First: All AI processing happens locally in your browser - no images uploaded to servers
• Confidence Scores: Each prediction includes a confidence percentage
• Real-time Processing: Powered by TensorFlow.js for fast, client-side inference
• Responsive Design: Optimized for all screen sizes and devices

Technical Stack

• Frontend: HTML5, CSS3, JavaScript (ES6+)
• AI Framework: TensorFlow.js
• Model: Custom Teachable Machine image classification model
• PWA: Service Worker for offline functionality and caching
• Deployment: GitHub Pages

How to Use

1. Initial Loading

• Open the application and wait for the custom AI model to load
• Status message will show "Loading custom AI model... Please wait"
• Once loaded: "Custom AI model loaded! Ready to analyze images."
• Camera and Gallery buttons will become active

2. Camera Mode

• Click Camera button
• Grant camera permissions when prompted
• Point camera at a person's face
• Click Analyze to capture and analyze the image
• Results will show the detected class name with confidence score
• Use Stop to exit camera mode or Clear to reset

3. Gallery Upload

• Click Gallery button
• Select an image file (JPEG, PNG, WebP supported)
• Image will automatically be analyzed after loading
• Results display immediately with class identification
• Use Clear to reset and try another image

4. Understanding Results

Results display:

• Class Name: Detected class from trained classes
• Confidence Score: Percentage indicating model certainty
• Alternative Predictions: Other possibilities with lower confidence scores
• Target Icon: Indicates custom model prediction

Development

Local Development

# Clone the repository
git clone https://github.com/anshulkhare7/ai-image-classifier.git
cd ai-image-classifier

# Serve locally
Simply open the index.html file in browser.

Project Structure

├── index.html          # Complete application (HTML + CSS + JavaScript)
├── manifest.json       # PWA configuration
├── sw.js              # Service worker for offline functionality
├── icons/             # PWA icons (192x192, 512x512)
├── CHANGELOG.md       # Version history
├── RELEASE_NAMES.md   # River-based release naming system
└── README.md          # This file

Version History

• v2.1.0 - Betwa (Current) - Updated custom Teachable Machine recognition model
• v2.0.0 - Yamuna - Custom Teachable Machine celebrity recognition model
• v1.0.0 - Ganga - Initial PWA release with multiple AI models (COCO-SSD, FaceMesh, MobileNet)

Release Naming

This project follows an Indian river naming convention for releases, starting with sacred rivers:

• v1.0.0 - Ganga (The most sacred river)
• v2.1.0 - Betwa (Sacred tributary)
• v2.0.0 - Yamuna (Sacred tributary of Ganga)
• Future: Saraswati, Narmada, Godavari, Kaveri...

Model Limitations

Current Model Constraints:

• Limited to 3 classes: Only trained on a specific set of classes
• Small dataset: 30 images per class (90 total training images)
• Lighting conditions: Performance may vary with different lighting
• Face angles: Works best with frontal face views
• Image quality: Higher resolution images generally yield better results

Future Enhancements

• Expand training dataset with more images
• Add more classes
• Improve model accuracy with data augmentation
• Add prediction confidence visualization
• Model versioning and easy switching between different trained models

Privacy & Data

• No data collection: Images are processed entirely on your device
• No server uploads: All AI inference happens in your browser
• No tracking: No analytics or user behavior tracking
• Open source: Complete transparency in code and functionality

Browser Compatibility

• Chrome 88+ (recommended)
• Firefox 85+
• Safari 14+
• Edge 88+
• Mobile browsers with camera support

Contributing

1. Fork the repository
2. Create a feature branch (git checkout -b feature/amazing-feature)
3. Commit your changes (git commit -m 'Add amazing feature')
4. Push to branch (git push origin feature/amazing-feature)
5. Open a Pull Request

License

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

Acknowledgments

• AI Development: Built with assistance from Claude Code by Anthropic
• Model Training: Teachable Machine by Google
• AI Framework: TensorFlow.js
• Icons: Emoji icons for consistent cross-platform display

---

Current Version: v2.1.0 - Betwa | Model: Custome Image Recognition | Classes: 3 trained classes