Sign Language Translator

A real-time sign language translation application using computer vision and deep learning.

Features

• Real-time sign language detection and translation
• Support for American Sign Language (ASL) alphabet
• Multiple operation modes:
  • Inference: Translate signs in real-time
  • Training: Train the model with your own data
  • Data Collection: Collect training data for custom signs
• Configurable settings for different environments and use cases

Requirements

• Python 3.7+
• TensorFlow 2.5+
• OpenCV 4.5+
• MediaPipe 0.8.10+
• NumPy 1.19+

Installation

1. Clone this repository
2. Install dependencies:

   pip install -r requirement.txt
   

3. Run the application:

   python main.py
   

Usage

Basic Usage

Run the application in inference mode (default):

python main.py

Command Line Arguments

• --model: Path to the trained model (default: models/sign_language_model.h5)
• --camera: Camera index to use (default: 0)
• --mode: Operation mode (inference, training, data_collection)
• --dataset: Dataset directory for training or collection
• --debug: Enable debug mode with visualization
• --subtitle_size: Size multiplier for subtitles
• --word_timeout: Timeout in seconds to form words

Examples

Collect training data:

python main.py --mode data_collection --dataset my_dataset

Train the model:

python main.py --mode training --dataset my_dataset --model my_model.h5

Run inference with debug visualization:

python main.py --debug

Controls

• Press ESC to exit the application
• Press c to clear the current text

Project Structure

• main.py: Main application entry point
• sign_detector.py: Core sign detection functionality
• utils/: Utility modules
  • config.py: Configuration management
  • logger.py: Logging setup
  • landmark_processor.py: Hand landmark processing
• models/: Directory for trained models
• dataset/: Default directory for training data

License

This project is licensed under the MIT License - see the LICENSE file for details.