Research on Automatic Recognition Method of Icon Style

Abstract

Icon is an important element in human-computer interaction, and icon style is the most intuitive visual expression of icon design. Aiming at the problem of material classification in the icon style design process, this paper proposes an icon style recognition method based on deep learning. This paper first established the icon style dataset, and then used Visual Geometry Group Network (VGGNet), AlexNet and self-built neural network for training. The results show that the accuracy of the trained icon style recognition model is up to 100%. In addition, convolved features were visualized for explaining the recognition progress. This method can help designers quickly collect and filter a large number of icons of the same type, sequentially improving and accelerating the icon design process.