Real-time Object Detection and Voice Labeling for Enhanced Accessibility and Visual Interaction
Matta Swathi, Ramala Supraja, Malavathu Lakshmi Prasanna, Shaik Sameer, Guntaka Rama Krishna Reddy · Advances in computer science research · 2024
This work introduces a new approach to real-time object recognition using YOLO Version 7, an advanced system capable of real-time object detection in images, videos, as well as live webcam feeds.Unlike traditional methods, this system verbally discusses everything it finds, including the object's name and the accuracy and confidence levels of the algorithm.Apart from enhancing accessibility, computers may also be leveraged to develop educational and engaging resources.Using the MS COCO dataset and a pre-trained model, YOLO Version 7 ensures accurate and speedy object recognition, even for sma ll objects.By using the speed and precision of the system, the initiative aims to make information less intimidating and engaging, particularly for individuals with visual impairments.The dataset ensures comprehensive evaluations with 118,287 training shots, 5,000 validating images, and 20,288 assessment images spanning 80 object classes.The following are the advantages of the proposed method: speed, accuracy, increased visual interactions, faster and less interference, flexibility in all situations, accurate and quick item recognition, and improved handling of small objects.The solution gathers data from several sources, including cameras and picture/video files, and recognizes objects using the YOLO Version 7 algorithm.