Modelo de clasificación de ímagenes basado en multiview learning

Abstract

This work proposes and develops an image classification model based on Multiview Learning (MVL). Its scope focuses on model generation and validation to improve the accuracy and relevance of content visualization systems. The project integrates various image views—convolutional neural networks, metadata analysis, object detection, and user feedback—to provide more robust, contextualized, and meaningful classifications. The case study focuses on images related to the armed conflict in Colombia, given the cultural, semantic, and historical complexity that this type of material entails. In this way, the model not only addresses a technical challenge but also a social one, contributing to a more accurate interpretation of sensitive content in immersive environments. Regarding its importance, the project contributes to methodological innovation by overcoming the limitations of traditional image classification with a one-dimensional approach by integrating multiple analytical perspectives. On a practical level, by enhancing the relevance and coherence of visual content in applications such as digital museums, augmented/virtual reality educational environments, and interactive cultural platforms. And on the level of academic and social impact, by opening up the possibility of designing systems that are more sensitive to the cultural and historical context, strengthening memory building and the user experience. The results achieved, with accuracy metrics above 80% and a high level of user satisfaction in terms of adaptability and customization, demonstrate that the proposed model is a significant contribution both to research on classification issues and to practical implementation in culturally and socially relevant visualization systems.