Intelligent Assistant System for Graphic Design Creation Process: Research on Creative Inspiration Extraction and Visualization Tools

Abstract

In graphic design, words, images, and creativity are combined to create a primarily "visual" medium for communication and expression. Design thinking and a variety of methodologies are also utilized to recreate the symbols, images, and words that are used in visual art to communicate ideas and information. When choosing design elements, a designer must make sure that the elements' contexts match the semantics provided by the product image and that the elements' visual styles such as their colors, forms, and sizes are perfectly harmonious. To address these challenges, a methodology for developing an intelligent assistant for graphic design creation is described in this manuscript. The images are collected fromImp1k dataset. The collected data are fed to pre-processing. Designs with skewed aspect ratios, few numbers of elements, or outliers within their design class were filtered out using the Adaptive Multi-scale Improved Differential Filter  during pre-processing. The pre-processed data is then sent to Multi-scale Hypergraph-based Feature Alignment Network for classification. The image design categories infographics, mobile-UI, movie posters and webpages are successfully classified using MHFAN. The Elk Herd Optimizer (EHO) is used to optimize the weight parameters of MHFAN. The proposed IAGDC-MHFAN-EHO method is executed on the Python working platform. Performance metrics like F1-score, accuracy sensitivity , precision and computation time are examined. The gained outcomes of the proposed IAGDC-MHFAN-EHO method attains higher accuracy of 16.71%, 18.82%, and 17.93%, higher sensitivity of 16.37%, 12.25%, and 18.51%  and higher precision of 14.93%, 16.79%, and 18.18%. The proposed IAGDC-MHFAN-EHO method is contracted with the  present technique  like DoGDAS-AI-DL, TADD-AIAD, and DAIAFDS-StyleGAN2 models respectively.