Telugu Braille Character Detection Using Deep Learning

Braille is a tactile writing system that visually impaired individuals use for reading and writing. While Braille has been adapted for many languages around the world, including major Indian languages like Hindi, robust Optical Character Recognition (OCR) systems are lacking for detecting Braille characters in Telugu, one of the classical Dravidian languages. This work presents a novel deep-learning approach for automatically recognising Telugu Braille characters from images. The proposed method involves preprocessing Braille images using techniques like principal component analysis for alignment and filtering for enhancement, followed by training a lightweight Convolutional Neural Network (CNN) architecture with an inverted residual block module on a data set of Telugu Braille images. The deep learning model achieves state-of-the-art accuracy in recognizing Telugu Braille characters while offering practical inference speed suitable for real-time applications. This work can potentially facilitate more accessible reading and writing tools for the visually impaired Telugu community.