Real-Time Sign Language Recognition

This project implements a real-time American Sign Language (ASL) recognition system using a pretrained CNN model and MediaPipe hands detection. The system captures hand gestures through a webcam, predicts the corresponding ASL letter, and constructs sentences dynamically. It also includes speech synthesis to read the constructed sentence aloud.

---

Features

• Real-time hand gesture detection using MediaPipe Hands
• CNN-based ASL alphabet classification with a custom pretrained model (models/asl_cnn.pth)
• Sentence construction from sequential ASL letter predictions
• Speech synthesis to read the sentence aloud (using gTTS and pygame)
• Clear sentence functionality with a keypress
• GPU support for faster model inference if CUDA is available
• Logs predictions and sentences to a log file for debugging and training analysis

---

Getting Started

Prerequisites

• Python 3.8 or above
• GPU with CUDA support (optional but recommended for faster inference)
• Webcam

Python Packages

Install required dependencies:

pip install -r requirements.txt