Tensorflow based mobile application for indoor object identification for visually impaired

Abstract

Vision is considered as one of the primary senses of human in obtaining inputs from the neighboring environment. It plays a vital role in human perception in identifying the environment around him. People who are visually impaired face many challenges and difficulties while interacting with the surroundings in visualizing images and objects. Detecting and recognizing generic objects needs to be performed as accurately and efficiently as possible. Many tools and techniques have been identified as an assistant to those people and as a support in decreasing the complications in their lives. So, industry has created a variety of computer vision products and services by developing new electronic technologies for the visually impaired in order to overcome the difficulties. Mobile computer vision is identified as a promising technology to support the people who are visually impaired. Fresh and exciting possibilities are being introduced with the emerging technology. Deep learning methods have resulted in significant performance improvements in several application domains and as such several software frameworks have been developed to facilitate their implementation. Object identification is considered as a major research area that has been focused in deep learning domain. The proposed system is a research-basedapplication, which uses the advantage of Tensorflow deep learning framework. The frameworkcan be used in many scenarios andfor this system it’s used to detect and identify indoor objects for visually impaired people. Thousand sets of indoor objects data are being fed to the system and used for the object identification.Audio guidance is also provided by the developed system to convey the response. The main idea of the project is to provide a better and accurate system in identifying the objects for the people who are visually impaired.The system is evaluated in two approaches. User acceptance testing is done in order to identify the obstacles and difficulties that the visually impaired people get when using the proposed system. Object detection evaluation is carried out in order to identify the efficiency and the accuracy of the system