| Budget Amount *help |
¥3,770,000 (Direct Cost: ¥2,900,000、Indirect Cost: ¥870,000)
Fiscal Year 2012: ¥910,000 (Direct Cost: ¥700,000、Indirect Cost: ¥210,000)
Fiscal Year 2011: ¥1,430,000 (Direct Cost: ¥1,100,000、Indirect Cost: ¥330,000)
Fiscal Year 2010: ¥1,430,000 (Direct Cost: ¥1,100,000、Indirect Cost: ¥330,000)
|
|---|
| Research Abstract |
The aim of this research is to develop environment adaptive image recognition and object detection methods that can optimize their performance in individual environment. This research considered this task as the combination of two sub-tasks: (i) automatic gathering of training samples from videos, and (ii) adaptation of object detection and recognition system according to an environment. To obtain training samples automatically, retrospective tracking method was developed for gathering positive samples, and characteristics of mis-detections were used for gathering negative samples. These methods were combined with semi-supervised learning and online learning frameworks to adapt the detector for individual environment.
|
