Kansei modeling on visual impression from small datasets

Abstract

Large datasets are generally required for machine learning. In order to improve the efficiency of the system, our team proposes a new Kansei modeling method, which requires users to collect only a small dataset. Using our method, the small datasets can search and collect large datasets classified in detail. As a result, our method creates the well-tuned Kansei model only from small datasets without the trouble of collecting many datasets. © 2011 Springer-Verlag.