Hand Sign Recognition

This project is a Hand Sign Recognition System using MediaPipe's hand tracking capabilities to detect and recognize hand gestures via a webcam.

Features

• Train the system with custom hand signs.
• Store multiple samples per sign for better recognition accuracy.
• Predict hand signs in real-time based on trained data.

How It Works

1. Training Process

1. Enter a Hand Sign Name in the input field.
2. Click Train Now to set the hand sign name.
3. Click Start Training to begin capturing hand sign samples.
4. Click Stop Training when enough samples are captured.
5. Repeat for multiple hand signs.

2. Prediction Process

1. Click Predict to switch to prediction mode.
2. Show a trained hand sign to the camera.
3. The system compares the detected landmarks with stored samples.
4. The closest match is displayed on the screen.

Components

HTML Structure

• Input field for Hand Sign Name.
• Buttons for Train Now, Start Training, Stop Training, Predict.
• A canvas for displaying hand landmarks.
• A hidden video element for webcam input.
• A section to display trained hand signs.

JavaScript Logic

• Training Mode: Captures hand landmarks and stores them under the given sign name.
• Prediction Mode: Compares detected hand landmarks with stored samples using Euclidean Distance.
• MediaPipe Integration: Uses MediaPipe's Hands API to detect hand keypoints and draw them on a canvas.

Technologies Used

• HTML & CSS: UI structure and styling.
• JavaScript: Hand gesture recognition logic.
• MediaPipe Hands API: Real-time hand tracking.
• Bootstrap: Responsive design.

Installation

1. Clone this repository:

   git clone https://github.com/tron01/HandSignPrediction.git

2. Open index.html in a browser.

Future Improvements

• Implement persistent storage for trained signs.
• Improve recognition accuracy with ML models.
• Add an export/import feature for trained gestures.

License

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