Using Deep Learning to Estimate User Impressions of Designs for 3D Fabrication

  • Koichi TaguchiEmail author
  • Manabu Hashimoto
  • Kensuke Tobitani
  • Noriko Nagata
Conference paper
Part of the Springer Proceedings in Physics book series (SPPHY, volume 233)


This paper proposes a method for applying three typical human impressions directly into product designs to fabricate products using 3D printers. The method automatically estimates human impressions of the three-dimensional shape of an item in terms of three representative sensibilities: “hard–soft,” “flashy–sober,” and “stable–volatile.” This technique can be used for new 3D fabrication processes that reflect the designer’s intentions directly into the shapes of products. To estimate the impressions of the shape of an object, we need to draw strong correlations between impressions, which are psychological factors, and the aspects of the shape of the object, which are physical factors. The method uses deep learning effectively to address this issue. The object being evaluated is first converted to a set of images by photographing it from 20 surrounding directions. This image set is used as input data for deep learning with parameters of human impressions of the object as supervisory signals. In experiments, we used original dataset of three-dimensional objects of a car with assigned impressions that had been quantified using the semantic differential (SD) method. The correlation coefficients between impressions estimated using this method and the supervisory signals for all the datasets were about 0.70 for “hard–soft,” about 0.61 for “flashy–sober,” and about 0.67 for “stable–unstable.”



This research was partially supported by the Center of Innovation Program from the Japan Science and Technology Agency, JST.


  1. 1.
    K. Tobitani, S. Akizuki, K. Katahira, M. Hashimoto, N. Nagata, A comparison study on 3D features in terms of effective representation for impression of shape, in The 2nd International Conference on Digital Fabrication, No. 22 (2016)Google Scholar
  2. 2.
    K. Taguchi, K. Sasaki, M. Hashimoto, K. Tobitani, N. Nagata, A proposal of 3D local feature for estimating human’s impression factor to shape of object, in International Workshop on Advanced Image Technology (2017)Google Scholar
  3. 3.
    C.R. Qi, H. Su, M. Niesner, A. Dai, M. Yan, L.J. Guibas, Volumetric and multi-view CNNs for object classification on 3D data, in IEEE Conference on Computer Vision and Pattern Recognition (2016)Google Scholar
  4. 4.
    C.E. Osgood, G.J. Suci, P.H. Tannenbaum, The Measurement of Meaning (University of Illinois Press, Oxford, England, 1957)Google Scholar
  5. 5.
    D.P. Kingma, Adam: a method for stochastic optimization, in International Conference on Learning Representations (2015)Google Scholar
  6. 6.
    Z. Wu, S. Song, A. Khosla, F. Yu, L. Zhang, X. Tang, J. Xiao, 3D ShapeNets: a deep representation for volumetric shapes, in IEEE Conference on Computer Vision and Pattern Recognition (2016), pp. 1912–1920Google Scholar

Copyright information

© Springer Nature Singapore Pte Ltd. 2019

Authors and Affiliations

  • Koichi Taguchi
    • 1
    Email author
  • Manabu Hashimoto
    • 1
  • Kensuke Tobitani
    • 2
  • Noriko Nagata
    • 2
  1. 1.Chukyo UniversityNagoya-shiJapan
  2. 2.Kwansei Gakuin UniversitySanda-shiJapan

Personalised recommendations