Saibhossain/bangla-sign-language-assistant

Real-time Bangla Sign Language Recognition Assistant using MediaPipe, OpenCV, and ML models. Features a FastAPI backend, webcam-based hand tracking, speech feedback, and an interactive web interface. # 🇧🇩 Bangla Sign Language Assistant 🤟 This project is a real-time Bangla Sign Language recognition tool that uses a webcam to detect hand gestures, predict the corresponding Bangla alphabet, and provide audio-visual feedback to help users learn sign language. ## 🚀 Features - 📷 Real-time hand landmark detection using MediaPipe - 🔤 Bangla alphabet prediction using trained ML model - 🎧 Audio guidance using `gTTS` and `pyttsx3` - 🌐 Web-based frontend (HTML + Jinja2) - 🧠 Model trained with Scikit-learn + Transformers - 🗣️ NLP support with BNLP Toolkit ## 🛠 Technologies Used - `mediapipe`, `opencv-python`, `scikit-learn`, `transformers` - `fastapi`, `uvicorn`, `jinja2` for web backend - `joblib`, `gtts`, `pyttsx3`, `playsound3` for voice feedback - `bnlp_toolkit`, `bnnumerizer`, `pyspellchecker` for Bangla text processing ## 📂 Project Structure ```bash . ├── bangla_sign_server.py # FastAPI backend ├── templates/index.html # Web frontend ├── model/bangla_handsign_alphabets_classifier_rf1 # Pretrained sign classifier ├── Dataset/bangla_signs_dataset.json # Sign metadata ├── requirements.txt ``` ## 🖥️ How to Run ### Step 1: Clone and Install Dependencies ```bash git clone github.com cd bangla-sign-language-assistant pip install -r requirements.txt ``` ### Step 2: Start the Server ```bash python -m uvicorn bangla_sign_server:app --reload ``` ### Step 3: Open in Browser Go to: ```bash 127.0.0.1 ``` You’ll see a live webcam feed with Bangla alphabet prediction and audio feedback. ## 📜 License ### 🔒 `.gitignore` file ```gitignore __pycache__/ *.pyc *.pkl *.safetensors .env .DS_Store .idea/ .vscode/