SignOutLoud: Sign language Recognition

Abstract: In everyday life, deaf people face many problems during simple task. One of the most difficult tasks for them is to communicate with other people. However only 0.25% of the total population use and practice Indian Sign Language. The proposed sign language Recognitions system aims to bridge the gap between non hearing and hearing population. This system is developed using Tensorflow, Tensorflow Object Detection API, Single Shot Detector (SSD). SSD is very popular algorithms for object detections amongst the other algorithms. These algorithms also provide high accuracy. This system converts the sign into text. India being more diverse and having many regional dialects still hasn’t its own sign language. Overall accuracy of the model is 85%.

Keywords: Sign to text, Single Shot Detector, TensorFlow (TF), Tensorflow Object Detection API