Hand Detection and Sign Language Recognition

A real-time hand detection and sign language recognition system that can identify hand gestures for letters A-L using computer vision and machine learning.

Features

• Real-time Hand Detection: Uses MediaPipe-based hand tracking to detect hands in live video
• Data Collection: Automated system to collect training images for different hand gestures
• Sign Language Recognition: Trained model to recognize hand signs for letters A through L
• Live Prediction: Real-time classification of hand gestures with visual feedback

Project Structure

HandDetection/
├── dataCollection.py    # Script to collect training data
├── test.py             # Main application for real-time recognition
├── model/
│   ├── keras_model.h5  # Trained neural network model
│   └── labels.txt      # Class labels (A-L)
├── Data/               # Training data organized by letter
│   ├── A/
│   ├── B/
│   ├── C/
│   └── ... (through L)
└── README.md

Requirements

• Python 3.7+
• OpenCV
• CVZone
• NumPy
• TensorFlow/Keras
• Webcam

Installation

1. Clone this repository:

git clone https://github.com/yourusername/HandDetection.git
cd HandDetection

2. Install required packages:

pip install opencv-python
pip install cvzone
pip install numpy
pip install tensorflow

Usage

Data Collection

To collect training data for a specific letter:

1. Open dataCollection.py
2. Change the folder variable to the desired letter (e.g., "Data/A" for letter A)
3. Run the script:

python dataCollection.py

4. Position your hand in the camera view showing the desired letter sign
5. Press 's' to save images
6. Collect multiple images (recommended: 100+ per letter)

Real-time Recognition

To run the hand sign recognition:

python test.py

• Position your hand in front of the camera
• The system will detect your hand and predict the letter
• The predicted letter will be displayed on the screen

How It Works

Data Collection Process

1. Hand Detection: Uses CVZone's HandDetector to locate hands in the camera feed
2. Image Preprocessing:
   • Crops the hand region with offset for better detection
   • Resizes to 300x300 pixels
   • Centers the hand image on a white background
   • Maintains aspect ratio to prevent distortion
3. Data Storage: Saves preprocessed images with timestamps

Recognition Process

1. Real-time Detection: Continuously captures frames from webcam
2. Hand Preprocessing: Same preprocessing as data collection
3. Classification: Uses trained Keras model to predict the letter
4. Visual Feedback: Displays the predicted letter on the video feed

Model Information

• Architecture: CNN (Convolutional Neural Network)
• Input Size: 300x300x3 (RGB images)
• Classes: 12 (Letters A through L)
• Framework: TensorFlow/Keras

Controls

• Data Collection Mode: Press 's' to save an image
• Exit: Press any key to close the application

Tips for Better Recognition

1. Lighting: Ensure good lighting conditions
2. Background: Use a plain background for better hand detection
3. Hand Position: Keep your hand clearly visible and well-positioned
4. Consistency: Maintain consistent hand poses during data collection
5. Distance: Keep a consistent distance from the camera

Future Improvements

• Extend to full alphabet (A-Z)
• Add number recognition (0-9)
• Implement word and sentence recognition
• Add voice feedback
• Improve model accuracy with data augmentation
• Add gesture-to-text functionality

Contributing

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

License

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

Acknowledgments

• CVZone library for simplified computer vision operations
• MediaPipe for hand tracking technology
• OpenCV for computer vision functionality

Contact

If you have any questions or suggestions, feel free to open an issue or contact me.

---

Note: This project is designed for educational purposes and basic sign language recognition. For production use, consider collecting more diverse training data and implementing additional error handling.