420220809-eng/HandSpeak

Egyptian Sign Language Translator --- title: HandSpeak API emoji: 🤟 colorFrom: blue colorTo: green sdk: docker pinned: false --- # HandSpeak - Arabic Sign Language Recognition API This is a FastAPI-based REST API for Arabic Sign Language Recognition using GRU (Gated Recurrent Unit) neural networks and MediaPipe for hand landmark detection. ## API Endpoints - **GET /** - API information and available endpoints - **GET /api/health** - Health check endpoint - **GET /api/model-info** - Get model information and sign classes - **POST /api/predict** - Upload video for sign language prediction - **GET /api/docs** - Interactive API documentation (Swagger UI) ## Usage ### Health Check ```bash curl daniel-10-handspeak-api.hf.… ``` ### Predict Sign Language ```bash curl -X POST "daniel-10-handspeak-api.hf.…" \ -H "Content-Type: multipart/form-data" \ -F "file=@your_video.mp4" ``` ### Interactive Documentation Visit: daniel-10-handspeak-api.hf.… ## Model Information - **Architecture**: GRU (Gated Recurrent Unit) - **Input**: Video files (mp4, mov, avi, mkv, webm) - **Output**: Predicted Arabic sign with confidence score - **Supported Signs**: 24 Arabic sign language gestures ## Technical Stack - **Framework**: FastAPI - **Deep Learning**: TensorFlow/Keras - **Computer Vision**: MediaPipe, OpenCV - **Deployment**: Docker on Hugging Face Spaces ## Response Format ```json { "success": true, "predicted_sign": "السلام", "confidence": 0.9876, "confidence_percentage": 98.76, "meets_threshold": true, "threshold": 0.7, "top_3_predictions": { "السلام": 0.9876, "عليكم": 0.0089, "شكرا": 0.0023 }, "extraction_metadata": { "total_frames": 45, "frames_with_hands": 42, "sequence_length": 30 } } ``` ## License This project is for educational and research purposes.