Design of Visual Communication Effect Evaluation Method of Artworks Based on Machine Learning

Abstract

With the rapid development of computer and digital media technology, more and more visual artworks are created, preserved, and transmitted digitally, which has become an indispensable spiritual wealth in this era. However, the aesthetic evaluation of traditional visual artworks can only rely on artists or appreciation experts for qualitative description, and the results have great subjective uncertainty due to their different personal knowledge and experience. Therefore, computational aesthetics came into being, that is, using a computer model to assist in quantitative evaluation of visual artworks. These models have broad application prospects in the fields of aesthetic evaluation and correction, art style identification, and so on. Based on this, this paper proposes a design of visual communication effect evaluation method of artworks based on machine learning. First, some characteristic variables are constructed to quantify the balance, contrast, and harmony in icon design criteria, and these three common design criteria are transformed into a mathematical expression. Based on the powerful fitting, classification, and generalization capabilities of the SVM method, we choose it as our base model. Then, the artificial evaluation scores are regressed to the calculated characteristic variables to obtain the statistical linear regression models corresponding to the three design criteria. The experimental results show that the evaluation model and manual evaluation results can reach a significant correlation in the same dimension, which verifies the effectiveness of the model.