A Computer Vision-Enabled Smart Healthcare and Assistive Technology Framework for Urban Digital Environments to Support Elderly Individuals and People with Disabilities

This research study presents a computer vision-enabled smart healthcare and assistive technology framework that aims to contribute to support systems for the elderly and people with disabilities in urban digital environments. The presented system incorporates real-time identification and classification using deep learning models, which would facilitate accurate and efficient demographic recognition. The framework is a scalable and adaptive solution to smart city applications in the realm of accessibility, monitoring, and healthcare support. Image acquisition and preprocessing, demographic identification by using YOLOv5, feature extraction and classification through EfficientNet-B0, and its performance evaluation based on several key metrics are part of the framework. The framework was trained and evaluated using the publicly available “R-CNN Dataset” from Roboflow Universe, which contains 3279 annotated images categorized into children, elderly individuals, and persons with disabilities. The YOLOv5 model demonstrated strong detection performance, achieving a precision of 0.82, a recall of 0.78, and a mean average precision (mAP) of 0.85 at Intersection over Union (IoU) 0.5, with a [email protected]:0.95 of 0.67, indicating reliable object detection with slight performance drops at stricter IoU thresholds. For classification, EfficientNet-B0 achieved an overall test accuracy of 99.70%. The classification model was reliable, as the precision scores for the child, elderly, and with disability categories were 0.9900, 1.0000, and 1.0000, respectively, and the recall values were 1.0000, 0.9900, and 1.0000. The proposed framework plans to bridge the gap between real-time surveillance, assistive technology, and the area of healthcare applications. The system improves accessibility, safety, and monitoring for elderly people and people with disabilities by combining YOLOv5 for detection and EfficientNet-B0 classification. The high precision, recall, and classification accuracy of the system make it suitable for large-scale deployment in smart city infrastructure for a safer and more inclusive urban environment.