Chip, tool and algorithm vendors will form a new trade group called Embedded Vision Alliance. To be officially launched in May, the group will push computer vision to become the killer app for next-generation smartphones and media tablets.
“We believe embedded vision will have a huge impact in mobile devices in the next few years,” said Jeff Bier, an organizer of the group and principal of DSP consulting firm Berkeley Design Technology Inc. “It will proliferate in home A/V systems and consumers will come to expect it,” he said in a talk at the Linley Tech Mobile Conference in San Jose, California.
Plenty of technical challenges are ahead. Extracting useful information from videos and images requires “a hierarchy of processing elements [and] a diverse and dynamic set of algorithms,” said Bier.
Because that has worked well so far?
Those challenges will require highly parallel hardware with a mixture of programmable elements that can handle everything from simple to complex algorithms operating on everything from fast flows of pixels to complex software objects, he added.
Because that has worked well so far?
In addition to the hardware challenges, computer vision software is still in a nascent state.
“Our 30 year old architecture is perfect! It’s the hardware and software that need work.”
“There’s a lack of application development infrastructure today, so you need to be the system designer or the processor designer initially and expand out from there,” Bier said. The good news is “we have a good algorithmic foundation for computer vision [from 30 years of academic studies] that we can now get into mobile devices,” he added.
“Awesome giant algorithms stuffed into awesome giant hardware!”
Can you feel it?
Lusty venturists, myopic academics, and single-crop semiconductor farmers conspiring for the last of your gas money.
{ 4 comments… read them below or add one }
Craig,
Good post,
I am doing work in similar area; multi-sensor cameras.
We see same limitations you mention …. the existing processors and software models are not appropriate.
some info about us;
twitter.com/@camera_tech
http://www.linkedin.com/company/argus-microsystems
Why won’t the chip companies try a different way?
Craig,
I look at mobile chip platforms.
They are occupied by many challenges. Better ‘vision’ capabilities is still a relatively new area. I am seeing more job openings for “computational algorithm” staff … so looks like things are changing.
Also, chipmakers for dSLR cameras are “purist”. Very hard for them to abandon convention. The dSLR chips are dominated by vertically integrated Japanese vendors, so its not surprising. Unfortunately, this narrow view spills over to other applications…. surveillance, toys, robotics, machine vision and so on.
-Tek
some info about us;
twitter.com/@camera_tech
http://www.linkedin.com/company/argus-microsystems
Interesting. You are making me think. Innovation has to come from below to serve exactly what’s needed. If it trickles down from above then it’s a method with a different agenda.