| Presentation | 2021-09-17 A CNN model Using Neural Style Features for Predicting Aesthetic Impressions Score Distribution Yuya Ohagi, Kensuke Tobitani, Iori Tani, Sho Hashimoto, Noriko Nagata, |
|---|---|
| PDF Download Page |  PDF download Page Link |
| Abstract(in Japanese) | (See Japanese page) |
| Abstract(in English) | In this study, we propose a method for predicting the probability distribution of aesthetic impression scores considering individual differences in impression evaluations using a deep neural network. We adopted neural style features, which potentially have relationships with visual impressions as explanatory variables. Then, we constructed a convolutional neural network (CNN) that estimated the probability distribution of impression scores based on product images. Next, we visualized attention maps that represented image areas that contribute to impression scores by using Grad-CAM. We also conducted an impression evaluation experiment to relate individual impression scores to the image areas that each participant considered important. Finally, we confirmed the similarity among the image areas by comparing the attention maps and the experimental results. |
| Keyword(in Japanese) | (See Japanese page) |
| Keyword(in English) | kansei (affective) engineering / visual impression / impression estimation / visualization / deep learning |
| Paper # | MVE2021-14 |
| Date of Issue | 2021-09-10 (MVE) |
| Conference Information | |
| Committee | MVE |
|---|---|
| Conference Date | 2021/9/17(2days) |
| Place (in Japanese) | (See Japanese page) |
| Place (in English) | Online |
| Topics (in Japanese) | (See Japanese page) |
| Topics (in English) | |
| Chair | Masayuki Ihara(RIKEN) |
| Vice Chair | Kiyoshi Kiyokawa(NAIST) |
| Secretary | Kiyoshi Kiyokawa(Oosaka Inst. of Tech.) |
| Assistant | Naoya Isoyama(NAIST) / Takenori Hara(DNP) / Mitsuhiro Goto(NTT) |
| Paper Information | |
| Registration To | Technical Committee on Media Experience and Virtual Environment |
|---|---|
| Language | JPN |
| Title (in Japanese) | (See Japanese page) |
| Sub Title (in Japanese) | (See Japanese page) |
| Title (in English) | A CNN model Using Neural Style Features for Predicting Aesthetic Impressions Score Distribution |
| Sub Title (in English) | |
| Keyword(1) | kansei (affective) engineering |
| Keyword(2) | visual impression |
| Keyword(3) | impression estimation |
| Keyword(4) | visualization |
| Keyword(5) | deep learning |
| 1st Author's Name | Yuya Ohagi |
| 1st Author's Affiliation | Kwansei Gakuin University(Kwansei Gakuin Univ.) |
| 2nd Author's Name | Kensuke Tobitani |
| 2nd Author's Affiliation | University of Nagasaki(Univ. of Nagasaki) |
| 3rd Author's Name | Iori Tani |
| 3rd Author's Affiliation | Kobe University(Kobe Univ.) |
| 4th Author's Name | Sho Hashimoto |
| 4th Author's Affiliation | Seinan Gakuin University(Seinan Gakuin Univ.) |
| 5th Author's Name | Noriko Nagata |
| 5th Author's Affiliation | Kwansei Gakuin University(Kwansei Gakuin Univ.) |
| Date | 2021-09-17 |
| Paper # | MVE2021-14 |
| Volume (vol) | vol.121 |
| Number (no) | MVE-179 |
| Page | pp.pp.33-37(MVE), |
| #Pages | 5 |
| Date of Issue | 2021-09-10 (MVE) |