Intelligent Context-Aware Visual Assistance with Real-Time Object Detection Using Deep Learning

Farid Rolando H.M.; Raul Jersson V.R.; Acosta Ticse D.L. · 4th Cognitive Models and Artificial Intelligence Conference, AICCONF 2026 - Proceedings (2026) · DOI: 10.1109/aicconf69182.2026.11600692

Resumen

The integration of deep learning technologies into assistive systems has opened new pathways to enhance accessibility for visually impaired individuals. This paper presents a mobile application leveraging MobileNetV2 and TensorFlow Lite to perform real-time object detection and environment description, designed specifically for visually impaired users in Lima, Peru. The system provides audio feedback to describe detected objects and their spatial locations, enabling greater independence and safety in daily activities. The methodology employs customized datasets for urban and domestic scenarios, achieving 85.3% precision, 82.7% recall, and real-time processing at 30 FPS. A dedicated Peruvian banknote recognition module further extends system capabilities toward financial autonomy. User validation conducted at CERCIL using the Technology Acceptance Model (TAM) yielded mean scores above 4.6 out of 5.0, confirming strong acceptance and practical usability.

Instituciones: Universidad Peruana de Ciencias Aplicadas