Real-time Identification System Using YOLO and Isomorphic Object Identification for Visually Impaired People
Wenxuan Zhang, Yuji Watanabe · 2023
In this study, we first design and implement a real-time object identification system for visually impaired people using a lightweight deep learning YOLO. Then, we verify the identification performance of our system for three types of milk boxes, which are objects with almost the same shape but different patterns. As a result of testing various images, an average confidence score of 96% was achieved at a distance of up to 30 cm. In addition, images with large changes in brightness and contrast can be correctly identified.