VGG16 CNN based Braille Cell classifier model for Translation of Braille to Text.

Abstract

The visually impaired individuals can use only Braille documents as a medium of education. These Braille documents are scripted on metallic plates which are heavy and bulky in nature. Maintaining these documents are very cumbersome due to their wear and tear tendency in unsupported weather conditions. It is also a tedious task to transport them over the globe and chances of damaging the documents is more. The main purpose of the research is to translate the Braille documents to natural text. Once the document is text form, it can be distributed over the globe using world wide web and can be reproduced to Braille document whenever needed. As the document is in digital form, maintaining Braille document for latter reproducing will be very easy.

Extracting the Braille Character cell from the Braille document is the central theme of the research. In previous work the paper was published on segmentation and cropping of Braille cells using its physical properties was published. This paper presents the Deep learning based proposed model for classifying the Braille cells using VGG16 Convolution neural network model.