Recognize American Sign Language letters using deep learning!
This project trains a Convolutional Neural Network (CNN) to classify hand gestures from grayscale images into 29 classes — the English alphabet (A–Z), plus space and nothing. It supports training, webcam inference, and browser deployment via TensorFlow.js.
- 29-class classification: A–Z, space, nothing
- Trained on 87,000+ preprocessed hand gesture images
- Run predictions in real time using your webcam
- Exportable to TensorFlow.js for browser-based use
- Fully modular: dataset → training → model → deployment
Required packages:
tensorflow pandas numpy scikit-learn opencv-python matplotlib joblib
The dataset was sourced from Kaggle: ASL Alphabet Dataset and converted into a single .csv with pixel-normalized grayscale values for training.
A → 0, B → 1, ..., Z → 25, space → 26, nothing → 27
- Dataset by grassknoted on Kaggle
- Help of A.I. tools for brainstorming