Sign Language Detection

This project aims to detect hand gestures representing different letters and numbers in American Sign Language (ASL) using a deep learning model. It uses a pre-trained model to recognize gestures captured through a webcam and provides a real-time visual output along with speech feedback.

Features

• Real-time Hand Gesture Recognition: Detects ASL hand gestures in real time from a webcam feed.
• Speech Feedback: Provides spoken output of the detected gesture using pyttsx3.
• User-friendly Interface: Visual feedback with the predicted sign and accuracy shown on the screen.
• Support for ASL Characters: The model is capable of recognizing both numeric (0-9) and alphabetic (A-Z) gestures.

Requirements

Before running this project, you need to install the necessary dependencies.

1. Python

Ensure you have Python 3.6 or later installed. You can download Python from here.

2. Install Dependencies

You need to install the following Python packages:

• OpenCV for video capturing and image processing
• TensorFlow for deep learning model
• pyttsx3 for text-to-speech conversion

You can install all required dependencies by running:

pip install opencv-python tensorflow pyttsx3 numpy

3. Additional Tools

Ensure you have a working webcam for capturing the hand gestures.

---

Project Structure

Here’s a quick overview of the project structure:

Sign-Language-Detection/
├── model/
│   ├── sign-lang-detection.json
│   └── sign-lang-detection.h5
├── realtimedetection.py
├── README.md
└── requirements.txt

• model/sign-lang-detection.json: The architecture of the pre-trained model.
• model/sign-lang-detection.h5: The weights of the pre-trained model.
• realtimedetection.py: The main script to run the real-time hand gesture detection.
• README.md: The documentation for the project (this file).
• requirements.txt: A list of all Python packages required for the project.

---

How to Use

1. Clone the Repository

Clone the repository to your local machine:

git clone https://github.com/yourusername/Sign-Language-Detection.git

2. Set Up the Environment

Navigate to the project directory and install the required dependencies:

cd Sign-Language-Detection
pip install -r requirements.txt

3. Run the Real-Time Detection

To start the real-time hand gesture recognition, run the following command:

python realtimedetection.py

This will open the webcam window, where the system will start detecting the ASL hand gestures.

---

How It Works

1. Model Architecture: The model is a Convolutional Neural Network (CNN) trained to recognize 36 ASL characters (0-9 and A-Z). The model's architecture and weights are stored in sign-lang-detection.json and sign-lang-detection.h5 respectively.

2. Webcam Feed: The cv2.VideoCapture function is used to capture real-time video feed from the webcam.

3. Hand Gesture Preprocessing: The hand gestures are first cropped from the video feed, converted to grayscale, and resized to 48x48 pixels. These are then passed through the trained model.

4. Prediction: The model predicts the hand gesture and outputs the corresponding letter or number along with the confidence (accuracy).

5. Speech Output: The predicted label is spoken using the pyttsx3 library.

---

Example Output

Upon detecting a gesture, the following happens:

• A rectangle is drawn around the detected hand gesture.
• The predicted label (e.g., 'A', 'B', '1', etc.) is displayed at the top of the screen.
• The accuracy of the prediction is shown alongside the label.
• The predicted gesture is spoken aloud using text-to-speech.

---

Troubleshooting

Error: ModuleNotFoundError: No module named 'pyttsx3'

Solution: Install the pyttsx3 module using the following command:

pip install pyttsx3

Error: ValueError: Input 0 of layer is incompatible

Solution: Ensure that the image is resized correctly and matches the input shape expected by the model (e.g., 48x48 pixels for grayscale images).

---

Contributions

Feel free to contribute to this project. You can do so by:

• Reporting bugs or issues
• Submitting pull requests for improvements or features

---

Acknowledgements

• OpenCV: For video capture and image processing.
• TensorFlow: For building and training the deep learning model.
• pyttsx3: For text-to-speech conversion.
• The dataset used for training the model (if applicable).

---

This README provides an overview of the project, installation instructions, and usage guidelines. Feel free to modify it according to your specific requirements or additions you’ve made to the project.