b*******e
发帖数: 217 | | A***l
发帖数: 461 | 2 这题目太大了，细致一些？
【在 b*******e 的大作中提到】
: RT
| k**********g
发帖数: 989 | 3
https://www.google.com/atap/projecttango/#project
Computer vision without a mobile market is a niche field. Big players;
server-side; Citizenship required, etc.
Computer vision + mobile computing makes it eligible for consumer market.
However, until recently, mobile devices do not have enough computing power (
and energy/battery efficiency) to justify running serious computer vision
algorithms on the devices.
A company named Movidius claimed to have solved that issue at the silicon (
architecture) level. It is selected by Google for its Project Tango tablet,
a research platform for consumer-grade computer vision. http://www.movidius.com/
Currently the architecture of Movidius is kept under NDA, so nobody outside
the Project Tango university partners know what it is like. A few
architecture diagrams can be found on EETimes, but the diagram refers to
acronyms without explaining the principles of operation.
It is claimed that Movidius is more power efficient than mobile GPU
computing.
The company, and Google, apparently wants to see the chip going into every
consumer mobile device in the very near future.
【在 b*******e 的大作中提到】
: RT
| k**********g
发帖数: 989 | 4 Let me clarify a bit first.
Computer vision is a very broad field.
RGBD (depth sensing) is a problem that is very close to being completely
solved. This is first popularized by XBox Kinect and soon to be taken to the
next level by Movidius and Project Tango.
There are many other topics under computer vision. Some of them are
significantly tied to machine learning. |
|
