Vision Connect: A Smartphone Based Object Detection for Visually Impaired People

Abstract

Visual impairments have long afflicted common people. Visual aids have also been around for a long time to help blind people to overcome the challenges and difficulties they face in everyday life. The invention of the Braille script signifies the result of incessant effort to help them lead normal, self-dependent lives. Thus, a lot of research has been done and money has been spent on various techniques to improve on these existing technologies. The proposed technique is a system for visually impaired people which not only detects obstacles for them in real-time but also helps them find various objects of their need around them. This technique is an integration of a smartphone application with an object recognition system. The object recognition module is implemented using Tensorflow. The smartphone camera is used to detect objects and the feedback is given in the form of speech output. The accuracy of the model is 85% which is higher as compared to an existing smartphone based systems using YOLO.