Visionguard

Abstract

According to world health statistics 285 million out of 7.6 billion population suffers visual impairment; hence 4 out of 100 people are blind. Absence of vision restricts the mobility of a person to pronounced extent and hence there is a need to build an explicit device to conquer guiding aid to the prospect. This project attempts to make A visual and voice assistant for visually impaired people which can help them reach their destination and also help them to read sign boards using Computer Vision.This project proposes to build a prototype that performs real time object detection using deep neural network model, YOLOV3-tiny. Further the object, and the class of the object is prompted through speech stimulus to the blind person. Along with this we are augmenting a voice assistant for frequent requirements and utilities such as sending emails, getting information over internet, etc. This work uses a combination of YOLOv3-tiny on pretrained datasetto build rapid real time multi object detection for a compact, portable and minimal response time device construction. Several prototypes and models have been made keeping in mind the blind people having different usages such as object recognizer for blinds, visual aid for blinds, google lookup, etc. Among these, we want to create a computer vision assisted solution keeping in mind their needs and their movements. Computer Vision based solutions are emerging as one of the most promising options due to their affordability and accessibility. This project proposes a system for visually impaired people. The proposed project aims to create a wearable visual aid for visually impaired people