Development of an AI-based object detection tool for the visually impaired using Raspberry Pi and a camera. Case study: Griya Harapan Social Service Center for the Disabled, Cimahi, Jawa Barat

Abstract

The limitation of identifying Objects in the surrounding environment is a major problem for blind people, especially those with low vision (who cannot see at all), which impacts their independence and participation in socio-economic life. To address this problem, the Community Service Team of the Faculty of Informatics at Telkom University conducted community service activities funded by Diktisaintek to develop and implement an AI-based visual aid for the blind. This aid utilizes a Raspberry Pi, camera, and headset that can recognize approximately 80 objects. The implementation method includes observation at the Griya Harapan Difabel Social Service Center as a partner, technology design, socialization and training on its use, field assistance, and evaluation. The evaluation model was carried out by implementing direct interviews with users to determine the objects that can be detected by the aid when using the device. The implementation results showed that the recognition accuracy level reached 80%. The social impacts achieved include increased independence for people who are blind, reduced risk of accidents, and new opportunities to participate in socio-economic activities. Obstacles encountered include limited datasets, variations in initial user skills, and the relatively high cost of the device. However, there are still ample opportunities for development, such as integration with mobile applications, improving CNN accuracy, and adding GPS features. This program demonstrates that technological innovation can be a sustainable solution to support the well-being and independence of people with disabilities.