Sign Language Detection for Deaf and Dumb students using Deep learning: Dore Idioma

Dore Idioma, also known as a hand in Arabic and sign language in Albanian, will be a futuristic website for the deaf-dumb or deaf-mute people as it provides a solution for people who face difficulty speaking and hearing. According to various surveys and stats from different deaf-mute organizations, 63% [21] of the world's population is deaf-mute. To help such people, we devised the idea of Dore Idioma. The fundamental functionality of Dore Idioma is to convert sign language (hand gesture). As communication takes place between ordinary and deaf-mute people, therefore, it will be easy for everyone to communicate through Dore Idioma. It is creating a user-friendly application that helps the user to have an approach to the translator very quickly. We will use image labeling so that input is given via the user; the information is then compared with the pre-trained detection model, after which the detected sign is cross-referenced with the database, and the final output will be generated. Various other models/software has been created. Still, they were not efficient and user-friendly enough and implementing. Dore Idioma might help in practical and better communication amongst ordinary and deaf-mute people. The pre-trained model used to create Dore Idioma is MobileNet version 2 (Convolutional Neural Network). Our current model is working on an accuracy of 99% and a total loss of 0.38%.