Técnicas de visión artificial aplicadas a tecnologías asistivas para personas con discapacidad visual: una revisión sistemática de la literatura
TL;DR
A Systematic Literature Review of artificial intelligence models in the domain of text, color, and object recognition revealed that architectures based on Convolutional Neural Networks (CNN) and deep learning models are the most effective for processing visual information through images or real-time video without demanding substantial computational resources.
Abstract
Visual impairment constitutes one of the primary barriers to social inclusion, affecting over 285 million individuals experiencing low vision or blindness, while assistive devices are frequently costly or lack accessibility. Consequently, the present study conducted a Systematic Literature Review (SLR) with the main objective of identifying the most efficient artificial intelligence models in the domain of text, color, and object recognition—specifically concerning items such as fruits, banknotes, and facial expressions. This research encompasses articles published between 2020 and 2026, exploring prominent indexed scientific databases including Redalyc, the ACM Digital Library, Consensus, IEEE Xplore, Springer Nature Link, and Google Scholar. Throughout the search process, 46 articles were identified and analyzed, revealing that architectures based on Convolutional Neural Networks (CNN) and deep learning models are the most effective for processing visual information through images or real-time video without demanding substantial computational resources.
