Sign Language Digits Recognition 👋🔢

A deep learning model that accurately classifies hand sign images representing digits 0-9 using convolutional neural networks. Achieves over 95% accuracy on the Sign Language Digits Dataset.

🌟 Features

• CNN architecture with 4 convolutional layers
• Data preprocessing and normalization
• Comprehensive training and evaluation pipeline
• Visualization of training metrics and confusion matrix
• Model checkpointing and saving

📊 Results

• Training accuracy: 95.26%
• Test accuracy: 95.81%
• Precision: 96% (macro average)
• Recall: 96% (macro average)

🚀 Quick Start

# Clone the repository
git clone https://github.com/your-username/sign-language-digits-recognition.git
cd sign-language-digits-recognition

# Install dependencies
pip install -r requirements.txt

# Download dataset (place X.npy and Y.npy in data folder)
# Train the model
python src/train.py

# Evaluate the model
python src/evaluate.py

##📁 Project Structure

sign-language-digits-recognition/
├── data/           # Dataset files
├── src/            # Source code
├── notebooks/      # Jupyter notebooks for exploration
├── models/         # Saved models
├── results/        # Output visualizations and reports
└── scripts/        # Utility scripts

Installation

1. Clone the repository:

git clone https://github.com/your-username/sign-language-digits-recognition.git
cd sign-language-digits-recognition

2. Install dependencies:

pip install -r requirements.txt

3. Download the dataset (if not already present):

python scripts/download_data.py

Usage

1. Train the model:

python src/train.py

2. Evaluate the model:

python src/evaluate.py

3. Explore the data and model in the Jupyter notebook:

jupyter notebook notebooks/exploratory_analysis.ipynb

Results

The model achieves the following performance:

• Training accuracy: 95.26%
• Test accuracy: 95.81%

See the results directory for training curves, confusion matrix, and classification report.

Model Architecture

The CNN architecture consists of:

• 4 convolutional layers with max pooling and dropout
• 2 fully connected layers
• Output layer with softmax activation
• Total parameters: 164,618

📝 License

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

🤝 Contributing

Contributions, issues, and feature requests are welcome! Feel free to check the issues page.

👨‍💻 Author

Thogaruchesti Hemanth