Modeling Relationship between Visual Impression of Products and Their Graphical Features

  • Shimon Niwa
  • Toshikazu Kato
Part of the Communications in Computer and Information Science book series (CCIS, volume 373)


Consumers feel heavy burden to find products suited individual preferences owing to large quantity of them on the Web. Therefore, it is necessary to classify products based on subjective preferences. This paper describes the method to model the relationship between subjective visual impressions and objective graphical features through machine learning for each user. The way to describe the visual impression is to use adjectives used by the professional photographers. As graphical feature vectors, we compute Lab color histogram and SURF from product photos. We estimate the subjective visual impression of products and classify them by constructing classifiers for the impression groups using Random Forests. As a result of the experiment, the proposed method achieved 80.1% accuracy on average.


Modeling Subjective Visual Impression Machine Learning Graphical Features 


  1. 1.
    Tada, M., Kato, T.: Visual Impression Modeling Using SVM for Automatic Image Classification. In: 2006 IEEE International Conference on Systems, Man and Cybernetics, pp. 872–877. IEEE (2006)Google Scholar
  2. 2.
    Uesaka, S., Yasukawa, K., Kato, T.: Kansei Modeling on Visual Impression from Small Datasets. In: Stephanidis, C. (ed.) Posters, Part I, HCII 2011. CCIS, vol. 173, pp. 605–609. Springer, Heidelberg (2011)CrossRefGoogle Scholar
  3. 3.
    Spillman, L., Werner, J.S.: Visual Perception: The Neurophysiological Foundations. Academic Press (1990)Google Scholar
  4. 4.
    Bay, H., Tuytelaars, T., Van Gool, L.: SURF: Speeded up robust features. In: Leonardis, A., Bischof, H., Pinz, A. (eds.) ECCV 2006, Part I. LNCS, vol. 3951, pp. 404–417. Springer, Heidelberg (2006)CrossRefGoogle Scholar
  5. 5.
    Breiman, L.: Random Forests. Machine Learning 45, 5–32 (2001)CrossRefMATHGoogle Scholar

Copyright information

© Springer-Verlag Berlin Heidelberg 2013

Authors and Affiliations

  • Shimon Niwa
    • 1
  • Toshikazu Kato
    • 1
  1. 1.Faculty of Science and EngineeringChuo UniversityBunkyo-kuJapan

Personalised recommendations