| Presentation | 2016-12-15 [Invited Talk] Dynamic datasets for flexible pattern recognition Mitsuru Ambai, |
|---|---|
| PDF Download Page |  PDF download Page Link |
| Abstract(in Japanese) | (See Japanese page) |
| Abstract(in English) | Standard pattern recognition uses N labelled data to learn a mapping function that relates an input pattern to a pre-defined label. Deep learning made great progress on improving performances of such standard pattern recognition tasks including multi-class classification, segmentation and object detection, etc. Software libraries of deep learning have now been widely used by not only researchers but engineers who are not familiar with machine learning. Can any problems of pattern recognition be solved if we collect a sufficient number of labelled data? If we had them, it might be true. Unfortunately, it is impossible to collect labelled data for particular types of pattern recognition tasks, e.g. inferring control signals of a robot arm from an image sequence. In such the case, we believe that the most promising approach is to learn the inferring function by interacting with a virtual environment. While we should call conventional dataset such as MNIST and ImageNet as static dataset, we should call the virtual enviroment as dynamic dataset. This talk discusses flexible pattern recognition that learns from the dynamic dataset to deal with difficulty of labelled data collection. |
| Keyword(in Japanese) | (See Japanese page) |
| Keyword(in English) | grand challenge / dataset |
| Paper # | PRMU2016-113 |
| Date of Issue | 2016-12-08 (PRMU) |
| Conference Information | |
| Committee | PRMU |
|---|---|
| Conference Date | 2016/12/15(2days) |
| Place (in Japanese) | (See Japanese page) |
| Place (in English) | |
| Topics (in Japanese) | (See Japanese page) |
| Topics (in English) | |
| Chair | Eisaku Maeda(NTT) |
| Vice Chair | Seiichi Uchida(Kyushu Univ.) / Hironobu Fujiyoshi(Chubu Univ.) |
| Secretary | Seiichi Uchida(Kyoto Univ.) / Hironobu Fujiyoshi(NTT) |
| Assistant | Masaki Oonishi(AIST) / Takuya Funatomi(NAIST) |
| Paper Information | |
| Registration To | Technical Committee on Pattern Recognition and Media Understanding |
|---|---|
| Language | JPN |
| Title (in Japanese) | (See Japanese page) |
| Sub Title (in Japanese) | (See Japanese page) |
| Title (in English) | [Invited Talk] Dynamic datasets for flexible pattern recognition |
| Sub Title (in English) | The 2nd Grand Challenge of PRMU |
| Keyword(1) | grand challenge |
| Keyword(2) | dataset |
| 1st Author's Name | Mitsuru Ambai |
| 1st Author's Affiliation | Denso IT Laboratory, Inc.(Denso IT Laboratory, Inc.) |
| Date | 2016-12-15 |
| Paper # | PRMU2016-113 |
| Volume (vol) | vol.116 |
| Number (no) | PRMU-366 |
| Page | pp.pp.11-11(PRMU), |
| #Pages | 1 |
| Date of Issue | 2016-12-08 (PRMU) |