| Presentation | 2019-07-19 A Note on Semantic Evaluation of Images Generated by Text-to-image Generative Adversarial Networks Rintaro Yanagi, Togo Ren, Takahiro Ogawa, Miki Haseyama, |
|---|---|
| PDF Download Page |  PDF download Page Link |
| Abstract(in Japanese) | (See Japanese page) |
| Abstract(in English) | Evaluating the quality of generated images from input sentences is important to verify the effectiveness of text-to-image Generative Adversarial Network (GAN) models. In the evaluation, generated images are evaluated in the view of ``Whether generated images are visually pleasant or not'' and ``Whether generated images represent the content of an input sentence or not''. There are various methods to evaluate the former evaluation but there is not much discussion about the later evaluation. We propose a new evaluation method to evaluate how much the generated images contain the semantic meaning of an input sentence, and verify the quality of generated images. In the proposed method, we generate a sentence from an image generated from an input sentence. We define similarity between the generated sentence and the input sentence as our evaluation value. By utilizing the generated sentence and the input sentence, we can verify the quality of the generated images quantitatively. |
| Keyword(in Japanese) | (See Japanese page) |
| Keyword(in English) | Generative adversarial network / Evaluation metrics / Image-to-text model / Text-to-image model |
| Paper # | IMQ2019-5,HIP2019-33 |
| Date of Issue | 2019-07-12 (IMQ, HIP) |
| Conference Information | |
| Committee | IMQ / HIP |
|---|---|
| Conference Date | 2019/7/19(1days) |
| Place (in Japanese) | (See Japanese page) |
| Place (in English) | Satellite Campus, Sapporo City University |
| Topics (in Japanese) | (See Japanese page) |
| Topics (in English) | Image Media Quality and Human Information Processing |
| Chair | Toshiya Nakaguchi(Chiba Univ.) / Miyuki Kamachi(Kogakuin Univ.) |
| Vice Chair | Mitsuru Maeda(Canon) / Kenya Uomori(Osaka Univ.) / Shuichi Sakamoto(Tohoku Univ.) / Yuji Wada(Ritsumeikan Univ.) |
| Secretary | Mitsuru Maeda(Shizuoka Univ.) / Kenya Uomori(Sony Semiconductor Solutions) / Shuichi Sakamoto(NICT) / Yuji Wada(NTT) |
| Assistant | Hiroaki Kudo(Nagoya Univ.) / Masaru Tsuchida(NTT) / Keita Hirai(Chiba Univ.) / Atsushi Wada(NICT) / Hidetoshi Kanaya(Ritsumeikan Univ.) / Yuki Yamada(Kyushu Univ.) |
| Paper Information | |
| Registration To | Technical Committee on Image Media Quality / Technical Committee on Human Information Processing |
|---|---|
| Language | JPN |
| Title (in Japanese) | (See Japanese page) |
| Sub Title (in Japanese) | (See Japanese page) |
| Title (in English) | A Note on Semantic Evaluation of Images Generated by Text-to-image Generative Adversarial Networks |
| Sub Title (in English) | |
| Keyword(1) | Generative adversarial network |
| Keyword(2) | Evaluation metrics |
| Keyword(3) | Image-to-text model |
| Keyword(4) | Text-to-image model |
| 1st Author's Name | Rintaro Yanagi |
| 1st Author's Affiliation | Hokkaido University(Hokkaido Univ.) |
| 2nd Author's Name | Togo Ren |
| 2nd Author's Affiliation | Hokkaido University(Hokkaido Univ.) |
| 3rd Author's Name | Takahiro Ogawa |
| 3rd Author's Affiliation | Hokkaido University(Hokkaido Univ.) |
| 4th Author's Name | Miki Haseyama |
| 4th Author's Affiliation | Hokkaido University(Hokkaido Univ.) |
| Date | 2019-07-19 |
| Paper # | IMQ2019-5,HIP2019-33 |
| Volume (vol) | vol.119 |
| Number (no) | IMQ-131,HIP-132 |
| Page | pp.pp.21-24(IMQ), pp.21-24(HIP), |
| #Pages | 4 |
| Date of Issue | 2019-07-12 (IMQ, HIP) |