Evaluate the Brand Style of Automated Design Product by Machine Learning

Automatic design technology is becoming mature and applied to generate various new concept designs. However, with many new concept designs, it is necessary to select the optimal plan effectively. Therefore, we proposed an evaluation method that combined design automation-generated techniques and t-distributed stochastic neighbor embedding (t-SNE) of machine learning. In the design experiment, we used automated design to generate 200 samples with the image generation of the 2D concept, sample management to build the correlation between 2D and 3D, and complete the 3D model. Finally, we used the brand style classification system to evaluate the sample. In the existing classification system, two computer mouse brand styles were classified: Mad Catz and Genius brand style. The system was based on the t-SNE of machine learning and data visualization to predict generated samples. The generated computer mice of design automation were evaluated effectively. Moreover, the brand style presented by the appearance design of the new mouse generated by design automation conformed to the design assumption of the original brand style.