IoT based Object Detection using Deep Learning Technique for Visually Impaired People
Kasarapu Ramani, Syed Nazia Parveen, Taj, M. Sai Maduvani, M. Bhavitha · Zenodo (CERN European Organization for Nuclear Research) · 2021
Vision is one of the most significant human senses among all the human senses present, and it assumes a crucial job in the understanding of the surrounding environment. Visually impaired people find it difficult to move around without any supervision. Hence, objects knowledge helps Visually Impaired People (VIP) in their navigation and facilitates their daily life. A Deep learning model dealing with object classification has been proposed which may be helpful to Visually Impaired People (VIP) when they navigate in natural indoor and outdoor environments. Indeed, the considered objects are supposed to support such elements of the navigation as the VIP's localization, obstacle detection, monitoring of the mobility progress. The algorithm used here is SSD-Mobilenet Algorithm, the accuracy is 99% and the detection time is less than 2 seconds. Espeak is used to convert the text to audio format so that visually impaired people can easily know about object.