While voice-recognition software has enabled computer users to verbally interact with their PCs, Intel is looking to spur development of another of the computer's "senses" -- visual recognition. The company will post programming codes on its Web site to boost development of camera-enabled, vision-based apps.
While voice-recognition software has enabled computer users to verbally interact with their PCs, Intel is looking to spur development of another of the computer's "senses" -- visual recognition.
By the end of this week, Intel will place an open-source library of programming codes on its Web site in a bid to boost development of camera-enabled, vision-based applications.
Visual applications are already being implemented in a variety of applications today, from security systems that recognise facial features, to military targeting and tracking systems, to novelty programs such as games.
But Intel hopes that by fueling further innovation, vision-enabled computer systems will become far more common, said Wilf Pinfold, marketing manager for Intel's Microprocessor Research Lab.
Such systems typically require more powerful processors, potentially driving up demand for Intel's high-end chips.
Many potential uses
The possible implementations of visual-recognition technology are
widespread. In a factory environment, for instance, computers could help
identify defective parts on assembly lines, Pinfold said.
"You can have a camera that's looking at the conveyer belt as it goes by, identifying that there is an object, identifying the orientation it is in, and then identifying if it is a complete object and whether a piece of it is missing or something is broken," he said.
Pairing computer-vision with voice-recognition programs could also help eliminate needless errors.
"Imagine you're talking to a computer that's taking dictation and someone walks in your office and you turn your head to chat with them. The computer recognises that you're not talking to it anymore, and stops taking dictation," Pinfold said.
In another example, a speaker could be making a presentation using a computer projection on a large screen where the use of a typical mouse would be inconvenient. Using a "visually aware" computer, the speaker could use hand gestures to open or close files or move objects around on the screen.
Computers could even be designed to recognise sign language, Pinfold said.
Feedback will yield better processors
Intel's Microprocessor Research Lab sees vision-enabling software as the
next logical progression in application development.
"We look at advance applications from the perspective of the impact that they'll have on the design of the platform in the future," Pinfold said. "We think that we've got a core of library functions that are a good foundation for researchers to build on and to grow a fully fledged vision library."
By hosting the open-source library, Intel hopes to gain valuable insight that it can use in the development of future processors.
"We'd like to understand what things people can do with vision software," Pinfold said. "So the feedback is we get to see how the library gets used, and from that knowledge we can design better processors."
While the open-source library is intended to encourage research, some of the coding could have immediate value in current applications, according to Intel. For example, the camera calibration functions in the software library will allow the use of a wide angle lens to capture a large field of view and correct for the lens distortion that produces unflattering large noses in video conferencing.
"We believe that the open-source availability of this library will accelerate computer vision research and ultimately hasten the day when computer vision can be used in consumer products," said Fred Pollack, director of Intel's Microprocessor Research Lab, in a statement. "Working with academia has allowed us to consolidate the best known computer vision technology and the latest research into this software library."
The open-source software release this week will include C source code for all of the library's functionality and a royalty-free redistribution license. The source code was developed and debugged using the Windows environment. A version of the library for Linux will be released before the end of the year.
Intel will facilitate the growth and maintenance of the library by accepting submissions of source-code content reviewed by a panel of experts from the computer vision community. Reviewers will include representatives from major vision laboratories including Jitendra Malik, of the University of California Berkeley; Takeo Kanade, of Carnegie Melon University; Pietro Perona, of the California Institute of Technology; Irfan Essa, of the Georgia Institute of Technology; Carlo Tomasi, of Stanford University; and Trevor Darell, of the Massachusetts Institute of Technology.
