+ Computer Vision · Flagship
SignEase: ASL Learning & Recognition Platform
Final-year project: an American Sign Language learning platform with real-time sign recognition (99.5%+ accuracy on alphabets and digits), tutorials, quizzes and an AI chatbot.
99.5%+alphabet and digit recognition accuracy
Source code is private Project Presentation↓
Overview
SignEase combines several deep learning models with a full web platform for learning American Sign Language. A React frontend and Django REST backend deliver alphabet, numeral and vocabulary tutorials, each followed by a quiz, with real-time webcam sign recognition, a wrong-sign filter, emotion detection, a progress dashboard and an LLM-based ASL chatbot. The digit and alphabet recognisers (MobileNetV2 transfer learning) reach 99.5%+ accuracy.
Highlights
- 99.5%+ accuracy on ASL alphabet (A-Z) and digit (0-9) recognition with MobileNetV2 transfer learning
- 3 integrated AI models: pose prediction, an LLM-based chatbot and alphabet/digit recognition, plus a video-level pipeline
- Emotion detection model with 80%+ accuracy
- Wrong-sign filter that validates signs before accepting them
- Full learning platform: tutorials, quizzes, progress dashboard and reviews (React + Django REST)
Tech stack
Gallery
