Bangla Sign Language Translator
Project Overview
The Bangla Sign Language Translator is an innovative web application that bridges the communication gap for the deaf and hard-of-hearing community in Bangladesh. Using advanced machine learning techniques, the application can recognize and translate Bangla sign language gestures into written Bangla text in real-time. The system was trained on a comprehensive dataset of 45 classes (9,000 images, augmented to 36,000) covering 10 Bangla digits and 35 Bangla characters. I experimented with various models including MobileNetV2, ResNet, CustomCNN, and Teachable Machine to achieve the highest possible accuracy. Beyond translation, the application also serves as an educational tool, displaying corresponding ideal sign images for result verification and including tutorials and curated playlists to help users learn Bangla sign language.
Challenges
- Creating a robust dataset for Bangla sign language characters
- Developing a model with high accuracy for real-time translation
- Building an intuitive interface for both translation and learning
- Optimizing the model for web deployment
Solutions
- Collected and augmented a dataset of 9,000 images to 36,000 for comprehensive coverage
- Experimented with multiple models to achieve the best balance of accuracy and performance
- Designed a user-friendly interface with real-time translation and educational resources
- Optimized the model for web deployment using TensorFlow.js
Screenshots