/Sign-Language-Recognition

A multi-class classifier using TensorFlow on a sign language dataset, achieving 99% accuracy in training and over 95% accuracy in the validation dataset within 15 epochs, enhancing accessibility for the hearing impaired. used data from https://www.kaggle.com/datasets/datamunge/sign-language-mnist

Primary LanguageJupyter Notebook

Sign Language MNIST Classifier 🤟📚

This project implements a multi-class image classifier using TensorFlow to recognize American Sign Language (ASL) letters from the Sign Language MNIST dataset.

🔍 Overview

  • Model Type: Convolutional Neural Network (CNN)
  • Dataset: Sign Language MNIST (Kaggle)
  • Classes: 24 ASL alphabets (A–Y, excluding J and Z)
  • Training Accuracy: ~99%
  • Validation Accuracy: >95%
  • Epochs: 15
  • Framework: TensorFlow & Keras

📘 Source

This project is inspired by the DeepLearning.AI TensorFlow Developer Specialization on Coursera.

🚀 Results

Achieved high classification accuracy using a simple CNN model, demonstrating how deep learning can help improve accessibility for the hearing impaired.