Thursday, December 30, 2010

United States development recognition human vision system for super computer

Yale University School of engineering and applied science of Eugenio · card rousseil-15 November in Boston, Massachusetts, held at the high-performance embedded computing (HPEC) seminar that he and the research team has developed based on human visual system's super computer, and it was the development of similar computer than on the speed and energy efficiency is greatly improved.

People in the car, the vision and the brain's response capacity to easily help your understanding of roads and surrounding environment, comfortable driving.

However this appears to be human is very simple for quick identification of Visual environment's ability to implement on your computer, you need a large number of operations, it is also difficult to imitate human brain computer system to identify objects.

Card Rousse Gairo developed super computer system named "neural flow" (NeuFlow), its design is inspired by the human visual system, it can mimic the human visual system's neural network quickly identify themselves around the world.

It has used the New York University, Jahn · Le Gallery invention of complex Visual algorithms to run application suitable for artificial vision of large-scale neural network.

Card Rousse Gairo and Le library research focuses on allowing the system to their own judgment road conditions to guide the truck.

The system is also capable of real-time processing capacity of 10 million pixel image, you can identify the path common to various destinations (including other car, pedestrian, lights, sidewalk, etc.).

In addition, although the "neural flow" super computer system can be completed per second more than 100 billion times of operational tasks, but their power is very low, even less than mobile phone consumption.

And in the laboratory, has many image processor computer such as accomplish the same task, you need to consume 300 Watts of power.

The system has run faster while saving reasons, because researchers will super computer embedded in a chip.

And full-sized computer, "neural flow" system you want to compact. He said that the entire system size no more than a wallet-size, so it can be easily installed in the car or elsewhere.

Except for car navigation, "neural flow" system is used to improve the robots into harmful or difficult to access places the navigation capabilities; provides for fighters battle 360 degree Visual synthesis of the environment; or for dynamic monitoring on site, such as help find elderly falls, etc.

No comments:

Post a Comment