SignSpeak

A small data-collection and sign-language recognition project. This repository contains code to collect gesture data, train a model, and run realtime inference.

Repository layout (important files):

• data/ - collected .npy feature files (excluded from git by default)
• config/actions_config.py - action / label configuration
• models/ - saved models and encoders (model files are excluded by .gitignore)
• src/ - application code
  • collect_data.py / collect_data_gui.py - scripts to collect and label data
  • train_model.py - training script (produces a .h5 model)
  • realtime_inference.py - script for realtime camera inference
• TODO.md - project TODOs

Quick start

1. Create and activate a virtual environment (recommended):

python3 -m venv .venv
source .venv/bin/activate

2. Install dependencies:

pip install -r requirements.txt

3. Collect data (example):

python src/collect_data.py
# or use the GUI
python src/collect_data_gui.py

4. Train model:

python src/train_model.py

5. Run realtime inference (requires camera and a trained model in models/):

python src/realtime_inference.py

Notes

• Large binary files (models, .npy datasets) are ignored in .gitignore. Store them outside the repo or use Git LFS if you want them in the remote.
• If you need a GPU-enabled TensorFlow build, install the appropriate tensorflow package for your platform.

Contributing

Feel free to open issues or pull requests. If you'd like, I can add CI (GitHub Actions) to run linting and unit tests.

License

Add a LICENSE file if you want to set a license for this project.