Development of Impression Evaluation Models for Illustration Learning Support
Yuichiro Kinoshita, Tomofumi Nakano, Kentaro Go · 2016
Although several studies have focused on support systems for illustration learning, most studies discussed methodologies of learning and did not consider the quantification of the impression of drawn illustrations. This paper describes the construction of evaluation models that quantitatively evaluate the impression of illustration. We first collected 100 illustrations with records of their drawing process. The impressions of the collected illustrations were quantified by subjective evaluations. Illustration features related to the impressions were then selected from the drawing records. Based on the features and the quantification results, a set of impression evaluation models was finally constructed using neural networks. The performance tests of the constructed models demonstrated sufficient accuracy in evaluating the impression of illustrations.