| Presentation | 2022-12-05 Accuracy Improvement of Real Image Classification by Style Transfer Using Training Data Created from 3DCG Takeru Inoue, Youichi Tomita, Kouji Gakuta, Etsuji Yamada, Aoi Kariya, Masakazu Kinosada, Yujiro Kitaide, Ryusuke Miyamoto, |
|---|---|
| PDF Download Page |  PDF download Page Link |
| Abstract(in Japanese) | (See Japanese page) |
| Abstract(in English) | In general, in order to achieve sufficient accuracy by machine learning in practical applications, training data appropriate for the target that requires huge cost for creation is indispensable. A novel framework was proposed to reduce the cost required for dataset creation by using three-dimensional models of target object to generate two-dimensional images with labels for classification. However, existing work shows that classification accuracy of actual images becomes worse when a classifier is trained using rendered images from three-dimensional models. This paper proposes a training scheme that learns shapes of target objects more than the standard way to improve classification accuracy when a domain gap exists. Experimental results showed that the classification accuracy was improved by 11% when style transfer was applied to training data generating from three-dimensional models. |
| Keyword(in Japanese) | (See Japanese page) |
| Keyword(in English) | deep learning / computer graphics / style transfer / domain gap |
| Paper # | SIS2022-30 |
| Date of Issue | 2022-11-28 (SIS) |
| Conference Information | |
| Committee | SIS |
|---|---|
| Conference Date | 2022/12/5(2days) |
| Place (in Japanese) | (See Japanese page) |
| Place (in English) | |
| Topics (in Japanese) | (See Japanese page) |
| Topics (in English) | |
| Chair | Tomoaki Kimura(Kanagawa Inst. of Tech.) |
| Vice Chair | Naoto Sasaoka(Tottori Univ.) / Hakaru Tamukoh(Kyushu Inst. of Tech.) |
| Secretary | Naoto Sasaoka(NTT) / Hakaru Tamukoh(Kansai Univ.) |
| Assistant | Yoshiaki Makabe(Kanagawa Inst. of Tech.) / Yosuke Sugiura(Saitama Univ.) |
| Paper Information | |
| Registration To | Technical Committee on Smart Info-Media Systems |
|---|---|
| Language | JPN |
| Title (in Japanese) | (See Japanese page) |
| Sub Title (in Japanese) | (See Japanese page) |
| Title (in English) | Accuracy Improvement of Real Image Classification by Style Transfer Using Training Data Created from 3DCG |
| Sub Title (in English) | |
| Keyword(1) | deep learning |
| Keyword(2) | computer graphics |
| Keyword(3) | style transfer |
| Keyword(4) | domain gap |
| 1st Author's Name | Takeru Inoue |
| 1st Author's Affiliation | Meiji University(Meiji Univ.) |
| 2nd Author's Name | Youichi Tomita |
| 2nd Author's Affiliation | Meiji University(Meiji Univ.) |
| 3rd Author's Name | Kouji Gakuta |
| 3rd Author's Affiliation | Digital Printing & Solutions Co., Ltd.(Digital Printing & Solutions) |
| 4th Author's Name | Etsuji Yamada |
| 4th Author's Affiliation | Digital Printing & Solutions Co., Ltd.(Digital Printing & Solutions) |
| 5th Author's Name | Aoi Kariya |
| 5th Author's Affiliation | Digital Printing & Solutions Co., Ltd.(Digital Printing & Solutions) |
| 6th Author's Name | Masakazu Kinosada |
| 6th Author's Affiliation | Shinsei Printing Co., Ltd.(Shinsei Printing) |
| 7th Author's Name | Yujiro Kitaide |
| 7th Author's Affiliation | Shinsei Printing Co., Ltd.(Shinsei Printing) |
| 8th Author's Name | Ryusuke Miyamoto |
| 8th Author's Affiliation | Meiji University(Meiji Univ.) |
| Date | 2022-12-05 |
| Paper # | SIS2022-30 |
| Volume (vol) | vol.122 |
| Number (no) | SIS-293 |
| Page | pp.pp.38-43(SIS), |
| #Pages | 6 |
| Date of Issue | 2022-11-28 (SIS) |