Sign Language Interpreter

An Indian sign language interpreter that uses ml and ai to detect sign language and translate it to real text.

Comment

Indian Sign Language Interpreter

Hackathon: The Dev Challenge

Project Overview

This project aims to develop an AI-driven system that interprets and translates sign language gestures in real-time using a webcam. The system is designed to make education more accessible to deaf or hearing-impaired students by recognizing and converting Indian Sign Language (ISL) into readable text.

Key Features

  • Real-Time Gesture Recognition: The system captures hand gestures via a webcam, processes them, and predicts the corresponding ISL alphabet or number.
  • Indian Sign Language Support: The model supports recognition for the ISL alphabet (A-Z) and numbers (1-9).
  • User-Friendly Interface: Displays both the captured image and the predicted sign on the screen, making it easy to use.

Technologies Used

  • TensorFlow: For loading the pre-trained model and making predictions.
  • OpenCV: For capturing webcam images and displaying the video feed in real-time.
  • NumPy: For image processing and handling large datasets.

Key Challenges

  • Dataset Limitations: High-quality datasets for Indian Sign Language (ISL) are limited. The model may struggle to generalize and accurately predict gestures due to small or unbalanced datasets.
  • Variation in Gestures: ISL gestures can vary in terms of style, speed, and regional dialects. Capturing these variations is critical for improving the model's accuracy.
  • Real-Time Processing: Achieving real-time recognition without delays requires optimizing the model to ensure fast processing without compromising on performance.

Dataset

The model uses a dataset of hand gestures corresponding to ISL letters (A-Z) and numbers (1-9). The dataset can be:

  • A pre-trained model file (.h5).
  • A custom dataset captured using the webcam.

Sample Dataset

Future Work

  • Improve model accuracy by collecting a larger and more diverse dataset.
  • Implement continuous recognition for interpreting multiple signs in sequence.
  • Expand the system to support additional languages or dialects.

How to Run

  1. Download the project files along with the dataset.
  2. Run code.py to build and train the model.
  3. Run try.py to capture images using your webcam and make predictions.

Video Demonstration

Watch the demo on YouTube

Built With

Share this project:

Updates