CamVisors: A Low-Cost Smart Glass System Utilizing Computer Vision for the Blind

Overview

More than 250 million individuals have severe visual impairments or complete blindness. These individuals have difficulty independently navigating new and unfamiliar places for work, study, leisure, and health. White canes, service animals, and electronic services are helpful but limited aids to independent navigation. The synergy between computer science and computer vision can be capitalized to improve independent navigation for these individuals in their community spaces. However, this technology can be expensive and out of reach for the 80% of the world’s visually impaired who live in third-world countries and do not have the infrastructure, access, or capital for such aid. CamVisors was designed and developed as a consolidated solution. CamVisors requires only a cost-effective camera headset and a smartphone. CamVisors works through an application with an object detection model and obstacle avoidance algorithms, a wayfinding algorithm, a text-recognition model, audio infrastructure, and portability to microcontroller applications and smartphone apps. Trained using the COCO and Google Street View obstacle datasets, CamVisors aimed to outperform YOLOv7, the most used machine learning algorithm for object detection. CamVisors utilizes a node-based wayfinding algorithm to generate paths for obstacle avoidance. For text recognition, CamVisors utilized Tesseract, a commonly used engine for Optical Character Recognition. CamVisors was built on a Kivy framework to ensure cross-platform compatibility. CamVisors systems met all engineering goals. This system provides a comfortable, easy-to-use, and portable solution to assist visually impaired individuals with mobility, safety, and independence for less than $200. CamVisors performed exceptionally in testing, where object detection was 96% accurate and classification was 86% accurate. This project provides insight into how creatively deployed technologies can be utilized in novel engineering solutions at a low cost to overcome physical limitations and increase an individual’s independence in navigating daily spaces.

Competition history

• AJAS 2024 Category not listed

Related projects

Closest projects by meaning, across every fair and year in the corpus.

Browse more like this

• All Embedded Systems projects
• AJAS Embedded Systems projects
• Every AJAS 2024 project