Research on Semantic-driven Intelligent Color Design based on DNN

This paper aimed at proposing a method for semantic-driven intelligent color design as an efficient auxiliary for the designers, and doing research on the algorithm and realization. Considering the extreme complexity of mapping from semantic space to color schemes, this paper constructed a dynamic neural network containing three layers to simulate the process of choosing a color scheme, using k-fold cross validation to choose the termination criterion. Based on the training database, this paper developed a semantic-driven intelligent color design prototype system on the platform of SolidWorks with a seamless connection. It can assist the designers to do the color design with great efficiency at a high-level and improve the quality of products design.