Optimized RE-CNN-Based Multi-Objective Decision Framework for Visual Feature Evaluation in Computational Art Analysis and Interactive Media

Abstract

Recent developments in digital art preservation have facilitated the integration of deep learning algorithms into decision-support systems by means of computational methodologies. This study proposes a hybrid approach employing the Zebra Optimisation Algorithm to enhance the performance of Random Ensemble-based Convolutional Neural Network (RE-CNN) models for the assessment of industrial product designs. By translating aesthetic evaluations, emotional responses, and memorability ratings provided by human reviewers into quantifiable insights, the model generates valuable operational guidance for visual media in the context of mechanical engineering design processes. Enhanced predictive capabilities are achieved through the optimisation of RE-CNN parameters, allowing for more accurate analysis of diverse product aesthetics. The advisory system presented offers design engineers and product developers a perceptual image generation tool tailored to achieving specific design goals. Product image evaluation is conducted by human assessors utilising systems that integrate decision-tree methodologies with content analysis protocols, enabling the interpretation of colour properties alongside visual elements. The system contributes to automated design decision-making by offering data-informed support for product selection within commercial environments. The adopted research methodologies contribute to the optimisation of mechanical systems and user interfaces, ensuring that visual and emotional design considerations remain paramount. Within artificial intelligence frameworks, engineers are thus able to create specialised products while accelerating the development of mechanical design systems