| Budget Amount *help |
¥17,680,000 (Direct Cost: ¥13,600,000、Indirect Cost: ¥4,080,000)
Fiscal Year 2019: ¥5,590,000 (Direct Cost: ¥4,300,000、Indirect Cost: ¥1,290,000)
Fiscal Year 2018: ¥5,590,000 (Direct Cost: ¥4,300,000、Indirect Cost: ¥1,290,000)
Fiscal Year 2017: ¥6,500,000 (Direct Cost: ¥5,000,000、Indirect Cost: ¥1,500,000)
|
|---|
| Outline of Final Research Achievements |
In this research, we attempted to extend the conventional assistive system for the visually impaired, which allows a computer to recognize text and objects. The conventional systems have the implicit assumption that the target object or text to be recognized is right in front of the user. However, what a visually impaired person wants to know is not limited to be in front of the person. Therefore, we proposed an assistive system for "looking for something" using an omnidirectional camera as an example of a framework that can recognize objects whose location is unknown to the user. In this context, we confirmed that the omnidirectional camera is easier to use than the conventional smartphone-type camera through experiments with visually impaired people. We also proposed a regularization method called ShakeDrop for deep learning to improve the recognition accuracy of object recognition.
|
