You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
π§ HastSanket: Real-Time Indian Sign Language to Speech Translator
HastSanket is an AI-powered real-time translator that converts Indian Sign Language (ISL) gestures into voice and text, helping deaf and mute individuals communicate effortlessly in day-to-day interactions.
π Project Vision
In a fast-paced, voice-driven world, people with speech and hearing disabilities are often left unheard. HastSanket aims to create an inclusive communication bridge between ISL users and the rest of society, enabling barrier-free interaction in schools, hospitals, offices, and more.
πΈ Demo Screenshots
Landing Page:
Translate Page:
π Features
ποΈ ISL Gesture Recognition via webcam
π§ Machine Learning Model for dynamic gesture interpretation
π Real-time Voice Output using browser Speech Synthesis
π Live Text Captions for each detected sign
π» Modern UI with React + Tailwind CSS
π FastAPI WebSocket Backend for real-time communication
π§ Tech Stack
Frontend
Backend
ML/AI
React.js
FastAPI (Python)
TensorFlow (Keras)
Tailwind CSS
WebSockets
MediaPipe (hand tracking)
Web Speech API
OpenCV
NumPy, deque
π Setup & Run Locally
πΉ Frontend
cd client
npm install
npm start
πΉ Backend
cd python_script
python -m venv venv
venv\Scripts\activate # Windows
pip install -r requirements.txt
python server.py
Note: Ensure webcam is connected. Trained model and label map should exist in the model/ folder.
π§ How It Works
1. MediaPipe tracks hand landmarks in real-time.
2. Sequences of landmarks (40 frames) are passed to a trained LSTM model.
3. The model predicts the gesture, which is mapped to a label.
4. The result is sent via WebSocket to the React frontend.
5. The app displays text and speaks it using the browser's Text-to-Speech API.