Object and Person Recognition System for Blind

Abstract

Natural and man-made disasters present a wide range of challenges, which are amplified for people with disabilities. In most cases, assistive technology devices and services assist the disabled in performing daily tasks; yet, these are commonly restricted during and after disasters. One potential approach for people with vision impairment is a cost-effective wearable ‘Object detection’. The object detection provides environmental narratives while also establishing communication between the visually impaired user and a huge online knowledge base system capable of vocalizing narratives. This research is carried out as a proof-of-concept using Python OpenCV and a webcam. The study’s goal is to create assistive technologies that will help visually impaired people navigate their way out of potentially hazardous situations. Object detection is a branch of computer vision that finds instances of semantic objects in images and videos (by creating bounding boxes around them in our case). The annotated text can then be converted into voice responses, which provide the basic positions of the objects in the person’s/camera’s view. We will use our webcam to feed images to this trained model at 30 frames/sec, and we can set it to only process every other frame to speed things up.