ASL Sign Language Recognition Web App

A modern, real-time American Sign Language (ASL) letter recognition game and web application powered by deep learning and computer vision.

Overview

This project uses a webcam and a trained neural network to recognize ASL letters in real time. Users play a word game by signing each letter of a word, with instant feedback and scoring. The app is built with Python, Flask, TensorFlow/Keras, and MediaPipe.

Features

• Real-time ASL letter recognition using your webcam
• Interactive word game with categories (Animals, Food, Colors, Simple Words)
• Visual feedback, progress bar, and scoring
• Robust backend with smoothed prediction logic for reliability
• Clean, responsive web interface

Setup Instructions

1. Clone the repository:

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

2. Install dependencies:

   pip install -r requirements.txt

3. Download or train a model:
   • Place your trained model file (e.g., new_asl_model.h5) in the project root.
   • (Optional) Use the provided scripts to preprocess data and train your own model.
4. Run the web app:

   python web_game.py

5. Open your browser:
   • Go to http://localhost:5000 to play the game!

Usage

• Show your hand sign clearly in the camera.
• Hold the sign steady until it's detected.
• Complete the word by signing each letter in sequence.
• Get points for each completed word.

Contributing

Contributions are welcome! Please open an issue or submit a pull request for improvements, bug fixes, or new features.

License

This project is licensed under the MIT License. See LICENSE for details.