Real Time Bangla Sign Language Detection With TensorFlow Object Detection Using Deep Learning SSD

Abstract

The Image detection system has revolutionized the way modern technology work due to its multipurpose use. Detecting Bangla sign language is nothing but detecting visual data which uses an image detection system to predict the sign and show the possible output. Our research shows that the Bangla sign language detection system work by collecting user data to match those data with previously trained data to show the result. After going through different research papers on this field we have found that there is a lot of existing work in this field, especially the in English language and other major language in the world. However, there are very few papers that have worked on Bangla sign language detection and it is mandatory to put emphasis on the Bangla sign language detection system. In this preliminary research paper, we have proposed a Bangla sign language detection model to detect the visual data and predict the sign with percentage to display the result. In order to detect Bangla sign language, we have used the 36 standard Bangla sign alphabet such as (অ-o, ক-k). The model we have proposed detects the visual data using an image recognition system and matches that information with the 36 Bangla sign alphabets that we have trained then it shows the result with matching percentage and expected possible result. To achieve the result, we have used a deep learning approach that uses a convolutional neural network (CNN) to detect the image and match the information with the Bangla sign alphabet to recognize the Bangla sign language. The approach we took showed average accuracy of 90.30 percent.