| Project/Area Number |
16K16083
|
|---|
| Research Category |
Grant-in-Aid for Young Scientists (B)
|
| Allocation Type | Multi-year Fund |
|---|
| Research Field |
Perceptual information processing
|
|---|
| Research Institution | The University of Tokyo |
|---|
Principal Investigator |
Kawakami Rei 東京大学, 大学院情報理工学系研究科, 特任講師 (90591305)
|
|---|
| Project Period (FY) |
2016-04-01 – 2018-03-31
|
| Project Status |
Completed (Fiscal Year 2017)
|
| Budget Amount *help |
¥3,900,000 (Direct Cost: ¥3,000,000、Indirect Cost: ¥900,000)
Fiscal Year 2017: ¥1,300,000 (Direct Cost: ¥1,000,000、Indirect Cost: ¥300,000)
Fiscal Year 2016: ¥2,600,000 (Direct Cost: ¥2,000,000、Indirect Cost: ¥600,000)
|
|---|
| Keywords | 動き / 学習 / 深層 / 鳥 / 風力発電 / 物体検出 / 分類 / 検出 / 動画 / 深層学習 / 追跡 / ニューラルネットワーク / LSTM / CNN / Long Short-term Memory |
|---|
| Outline of Final Research Achievements |
In this project, we worked on the detection and classification of objects that are too small that they can only be recognized by their motion. With wide area surveillance of wild birds as a target application, we collected 4K resolution videos for sea eagles around the wind turbines and labeled bird trajectories and background parts for 768 GB videos. As a method of detection and classification, tracking using a correlation filter of deep feature and motion learning by convolutional LSTM leads to a method that identifies a bird by its motion patterns acquired by simultaneously tracking them. It achieved a performance improvement of 25.2% points from baselines that only use still-image features. The achievements appeared as several papers, and they will contribute to reduce the impact of wind turbines on the ecology of wild birds.
|
