| Presentation | 2018-03-06 Sonar2image: GAN-based night vision for fish monitoring Kento Shin, Kei Terayama, Katsunori Mizuno, Koji Tsuda, |
|---|---|
| PDF Download Page |  PDF download Page Link |
| Abstract(in Japanese) | (See Japanese page) |
| Abstract(in English) | Fish monitoring in an aquaculture farm is indispensable for managing fish growth and health status. However, it is not realistic for humans to monitor at night with an optical camera. Although SONAR(Sound navigation and ranging) can be used at night, the quality of its white and black images is too low to use practically. In this paper, we propose a method to generate realistic images from sonar images by using conditional Generative Adversarial Networks (cGAN), a kind of deep neural network model. cGAN learns the image-to-image translation between optical and sonar images. We created an image dataset of sardines ({it Sardinops melanostictus}) consisting of a large number of sonar and optic camera image pairs simultaneously recorded by a high precision imaging sonar ARIS and an underwater camera. Our experiment showed that the proposed model successfully generated realistic images from sonar ones with high quality. Our method enables nighttime monitoring, leading to more efficient farming and labor saving. |
| Keyword(in Japanese) | (See Japanese page) |
| Keyword(in English) | conditional GAN / image-to-image translation / fish farming / SONAR |
| Paper # | IBISML2017-102 |
| Date of Issue | 2018-02-26 (IBISML) |
| Conference Information | |
| Committee | IBISML |
|---|---|
| Conference Date | 2018/3/5(2days) |
| Place (in Japanese) | (See Japanese page) |
| Place (in English) | Nishijin Plaza, Kyushu University |
| Topics (in Japanese) | (See Japanese page) |
| Topics (in English) | Statisitical Mathematics, Machine Learning, Data Mining, etc. |
| Chair | Kenji Fukumizu(ISM) |
| Vice Chair | Masashi Sugiyama(Univ. of Tokyo) / Hisashi Kashima(Kyoto Univ.) |
| Secretary | Masashi Sugiyama(Nagoya Inst. of Tech.) / Hisashi Kashima(Univ. of Tokyo) |
| Assistant | Tomoharu Iwata(NTT) / Toshihiro Kamishima(AIST) |
| Paper Information | |
| Registration To | Technical Committee on Infomation-Based Induction Sciences and Machine Learning |
|---|---|
| Language | JPN |
| Title (in Japanese) | (See Japanese page) |
| Sub Title (in Japanese) | (See Japanese page) |
| Title (in English) | Sonar2image: GAN-based night vision for fish monitoring |
| Sub Title (in English) | |
| Keyword(1) | conditional GAN |
| Keyword(2) | image-to-image translation |
| Keyword(3) | fish farming |
| Keyword(4) | SONAR |
| 1st Author's Name | Kento Shin |
| 1st Author's Affiliation | University of Tokyo(Univ. of Tokyo) |
| 2nd Author's Name | Kei Terayama |
| 2nd Author's Affiliation | University of Tokyo(Univ. of Tokyo) |
| 3rd Author's Name | Katsunori Mizuno |
| 3rd Author's Affiliation | University of Tokyo(Univ. of Tokyo) |
| 4th Author's Name | Koji Tsuda |
| 4th Author's Affiliation | University of Tokyo(Univ. of Tokyo) |
| Date | 2018-03-06 |
| Paper # | IBISML2017-102 |
| Volume (vol) | vol.117 |
| Number (no) | IBISML-475 |
| Page | pp.pp.85-89(IBISML), |
| #Pages | 5 |
| Date of Issue | 2018-02-26 (IBISML) |