| Presentation | 2022-02-18 Training Data Generation Method for Object Recognition from Free Direction Masatomo Ozeki, Ayaka Kumeta, Tsukasa Kudo, |
|---|---|
| PDF Download Page |  PDF download Page Link |
| Abstract(in Japanese) | (See Japanese page) |
| Abstract(in English) | Object recognition utilizing deep learning is widely applied in various fields. On the other hand, the preparation of training data for deep learning often requires a large load and becomes an obstacle to applying it. In particular, to recognize the target object from free directions, training data from the various direction is required, and the load becomes higher. In this study, to efficiently generate such training data, we propose two approaches. One is a method of automatically extracting training data from continuously shot videos; the other is a method of automatically generating it by computer graphics (CG). Furthermore, we evaluate the training data generation efficiency and object recognition accuracy. And, it is shown that the former method is effective when the environment is specified; using both methods together is effective when the environment is not specified, namely various environments. |
| Keyword(in Japanese) | (See Japanese page) |
| Keyword(in English) | deep learning / object recognition / training data generation / computer graphics / CG / video |
| Paper # | SWIM2021-38 |
| Date of Issue | 2022-02-11 (SWIM) |
| Conference Information | |
| Committee | SWIM |
|---|---|
| Conference Date | 2022/2/18(1days) |
| Place (in Japanese) | (See Japanese page) |
| Place (in English) | Online |
| Topics (in Japanese) | (See Japanese page) |
| Topics (in English) | |
| Chair | Kenji Saotome(Hosei Univ.) |
| Vice Chair | Akihiro Hayashi(Shizuoka Inst. of Science and Tech.) |
| Secretary | Akihiro Hayashi(Tokyo Univ. of Science) |
| Assistant | Tsukasa Kudo(Shizuoka Inst. of Science and Tech.) / Kokichi Tsuji(Aichi Pref. Univ.) |
| Paper Information | |
| Registration To | Technical Committee on Software Interprise Modeling |
|---|---|
| Language | JPN |
| Title (in Japanese) | (See Japanese page) |
| Sub Title (in Japanese) | (See Japanese page) |
| Title (in English) | Training Data Generation Method for Object Recognition from Free Direction |
| Sub Title (in English) | Two Approaches Utilizing Video and CG |
| Keyword(1) | deep learning |
| Keyword(2) | object recognition |
| Keyword(3) | training data generation |
| Keyword(4) | computer graphics |
| Keyword(5) | CG |
| Keyword(6) | video |
| 1st Author's Name | Masatomo Ozeki |
| 1st Author's Affiliation | Shizuoka Institute of Science and Technology(SIST) |
| 2nd Author's Name | Ayaka Kumeta |
| 2nd Author's Affiliation | Shizuoka Institute of Science and Technology(SIST) |
| 3rd Author's Name | Tsukasa Kudo |
| 3rd Author's Affiliation | Shizuoka Institute of Science and Technology(SIST) |
| Date | 2022-02-18 |
| Paper # | SWIM2021-38 |
| Volume (vol) | vol.121 |
| Number (no) | SWIM-372 |
| Page | pp.pp.51-58(SWIM), |
| #Pages | 8 |
| Date of Issue | 2022-02-11 (SWIM) |