| Presentation | 2019-03-18 A Generative Self-Ensemble Approach to Simulated+Unsupervised Learning Yu Mitsuzumi, Go Irie, Atsushi Nakazawa, Akisato Kimura, |
|---|---|
| PDF Download Page |  PDF download Page Link |
| Abstract(in Japanese) | (See Japanese page) |
| Abstract(in English) | The simulated and unsupervised (S+U) learning framework is an effective approach in computer vision since it solves various recognition tasks without using labeled real images. Although both labeled synthetic and unlabeled real images are available, existing S+U learning methods use only the labeled synthetic images for training predictors (regression functions or classifiers), which may prevent from leveraging information of the target domain. In this paper, we propose a novel S+U learning approach that utilizes both synthetic and real images to improve the prediction performance in the real domain. Our method consists of a) unsupervised learning of one-to-many translations that can generate a wide variety of ``fake'' images from a single real image with preserving their labels, and b) semi-supervised self-ensemble learning that gains increased prediction accuracy by using both labeled synthetic and unlabeled real images. |
| Keyword(in Japanese) | (See Japanese page) |
| Keyword(in English) | Deep Learning / Generative Adversarial Nets / Semi-supervised Learning |
| Paper # | BioX2018-52,PRMU2018-156 |
| Date of Issue | 2019-03-10 (BioX, PRMU) |
| Conference Information | |
| Committee | PRMU / BioX |
|---|---|
| Conference Date | 2019/3/17(2days) |
| Place (in Japanese) | (See Japanese page) |
| Place (in English) | |
| Topics (in Japanese) | (See Japanese page) |
| Topics (in English) | |
| Chair | Shinichi Sato(NII) / Kazuhiko Sumi(AGU) |
| Vice Chair | Yoshihisa Ijiri(Omron) / Toru Tamaki(Hiroshima Univ.) / Hitoshi Imaoka(NEC) / Tetsushi Ohki(Shizuoka Univ.) |
| Secretary | Yoshihisa Ijiri(NEC) / Toru Tamaki(Osaka Univ.) / Hitoshi Imaoka(Fujitsu Labs.) / Tetsushi Ohki(Univ. of Electro-Comm.) |
| Assistant | Go Irie(NTT) / Yoshitaka Ushiku(Univ. of Tokyo) / Norihiro Okui(KDDI Research) / Daishi Watabe(Saitama Inst. of Tech.) |
| Paper Information | |
| Registration To | Technical Committee on Pattern Recognition and Media Understanding / Technical Committee on Biometrics |
|---|---|
| Language | JPN |
| Title (in Japanese) | (See Japanese page) |
| Sub Title (in Japanese) | (See Japanese page) |
| Title (in English) | A Generative Self-Ensemble Approach to Simulated+Unsupervised Learning |
| Sub Title (in English) | |
| Keyword(1) | Deep Learning |
| Keyword(2) | Generative Adversarial Nets |
| Keyword(3) | Semi-supervised Learning |
| 1st Author's Name | Yu Mitsuzumi |
| 1st Author's Affiliation | Kyoto University(Kyoto Univ.) |
| 2nd Author's Name | Go Irie |
| 2nd Author's Affiliation | Nippon Telegraph and Telephone Corporation(NTT) |
| 3rd Author's Name | Atsushi Nakazawa |
| 3rd Author's Affiliation | Kyoto University(Kyoto Univ.) |
| 4th Author's Name | Akisato Kimura |
| 4th Author's Affiliation | Nippon Telegraph and Telephone Corporation(NTT) |
| Date | 2019-03-18 |
| Paper # | BioX2018-52,PRMU2018-156 |
| Volume (vol) | vol.118 |
| Number (no) | BioX-512,PRMU-513 |
| Page | pp.pp.137-142(BioX), pp.137-142(PRMU), |
| #Pages | 6 |
| Date of Issue | 2019-03-10 (BioX, PRMU) |