(英) |
In this study, the facial images are classified into multiple clusters from the viewpoint of impressions of human faces. An impression evaluation experiment using the SD method was performed on the face images of 78 female models, and the factors that make up those face images were investigated. As a result, we were able to summarize three factors: competence, activity, and evaluation. Also, based on these factor structures, we were able to cluster face images in the impression space. In order to confirm the validity of this classification, average faces of face patterns belonging to each cluster are created, and verification is made as to whether visual similarity with the average face is observed in the same cluster. |