GAZIRU - Image recognition technology via mobile device [3:13]
As cameras have reduced in size and are now common features on mobile terminals, image resolution technology has also advanced rapidly. By shifting to broadband network, smart devices can now process massive data.
In this age, it can be said that mobile terminals have now acquired their own brain capacity, exhibiting an ability superior to human sight.
In response to various needs, NEC has delivered many solutions in image recognition.
Today, we introduce their latest image recognition service called "GAZIRU".
NEC 3rd Carrier Services Division Shigekazu Fukuzawa
By facing the smartphone towards this apartment, we get the data about its vacancy, rent and layout displayed in real-time.
Similarly, car data can also be retrieved which not only includes the model names but also information on special offers as well.
For food, by scanning it with a smartphone, we can get recipes and calorie information.
In the case of pasta, for example, pescatore and carbonara can be distinguished.
Gaziru can also distinguish between chicken curry and keema curry.
This technology can be utilized for videos as well as images.
For example, when the camera is held towards this commercial, the name of the actress is displayed.
Image recognition technology opens possibilities for a variety of services.
Data on cars or food can be retrieved from images in magazines.
For business, we can analyze market trends from images in SNS and link related ads.
So, how does image recognition technology work exactly?
NEC Central Research Laboratories Toshihiko Hiroaki
A mechanical device must first learn identifiers from a huge archive of related images for accurate recognition.
First, the required data goes through image processing. Images are outlined, and then other characteristics such as the number of vertical, horizontal, diagonal lines and color types are digitized. These factors are mapped and by entering differentiating lines between them develop the learning process. By creating this map for characteristics and then mapping a new image, the device can now recognize what the image is.
Image recognition technology is rapidly advancing, promising further unique applications in the near future.
(January 25, 2013)