Object Detection with Voice Feedback and Distance Estimation

Abstract

The rise of visual impairment is increasing worldwide. There are many people who have problems of visual impairment and blindness. A person with little or no vision usually relies on walking sticks, smart glasses, braille systems and other optical tools. However, to assist those who have near-sightedness or no vision at all, the visual world must be converted into an audio world, which can help them understand surrounding objects and their spatial locations. As a result, we propose to assist the visually impaired by providing a system that is the most practicable, simplistic and cost-effective. Object detection is an important technique based on computer vision that identifies objects in images and videos. In this project, we are detecting objects in real time, using You Only Look Once v3 (YOLO v3) which is a real time object detection algorithm that runs on a variation of an extremely complex Convolutional Neural Network (CNN) architecture called the Darknet, after object is detected, we estimate the distance from camera to object and using a text-to-speech conversion tool like gTTS, provide voice feedback to the user.