By Brian Dipert
Editor-In-Chief
Embedded Vision Alliance
Senior Analyst
BDTI
Hopefully by now you've had the opportunity to audition the keynote delivered by Gary Bradski, President and CEO of the OpenCV Foundation and Founder and CTO at Industrial Perception, at last month's Embedded Vision Alliance Member Summit. Bradski will also be the afternoon keynoter at next month's Embedded Vision Summit.
One of the key themes of Bradski's talk last month was the recent (a few weeks earlier, to be precise) incorporation of the non-profit OpenCV Foundation, and the transition of overall OpenCV (Open Source Computer Vision Library) project management from Willow Garage (Bradski's former employer, where he was senior scientist) to it. Willow Garage continues to be a leading OpenCV backer, with additional financial and other contributions coming from Google and NVIDIA. Jump to 20:45 in the below video for the first portion of Bradski's OpenCV Foundation discussion, and then to 40:45 for more information:
Subsequent to the July Member Summit, I emailed Bradski with a few follow-up questions. You'll find them below, along with Bradski's responses, unchanged by me save for minor grammatical corrections in a few instances (coupled with web links I've added for your additional-information purposes).
I've also created a site discussion forum thread related to this article, which I'll ask Bradski to monitor. I encourage you to post additional OpenCV Foundation comments and questions there.
Q: Is the GPU optimization work that NVIDIA is doing vendor- and product-generic, or proprietary to NVIDIA (and perhaps even to particular NVIDIA products)?
A: NVIDIA is creating optimized code in CUDA, in particular working on optimized code for Tegra. To the extent that others use CUDA, they'd directly benefit. But, all of it is open source, so anyone would benefit from seeing the code organized in a parallel fashion for data flow. Thus, CUDA should be much easier to turn into OpenCL than the C++ code. Also, NVIDIA has contributed many new algorithms to OpenCV. I'd be more than happy if someone wanted to finance a pure OpenCL effort, which should go pretty fast on the code that has been converted to CUDA. OpenCV.org is the perfect place to do this.
Q: Is there a summary of the OpenCV Foundation's mission statement, current status and plans published anywhere?
A: The filed mission statement is, "To advance all aspects of machine and computer vision: develop and improve algorithms, educate practitioners and users, increase the number of uses and users, enable free and/or commercial deployment of computer vision applications in all areas while providing an industrial quality open and free infrastructure to help achieve these goals."
Q: Is there any other recent OpenCV Foundation news that I should consider mentioning in my writeup?
A: Khronos work on a low level standard vision API is ongoing and people from the OpenCV team are closely involved in the work [editor note: Bradski discusses Khronos' OpenVL API beginning at 39:25 in the above video].
Also, note some of the OpenCV websites:
- http://www.opencv.org is the user site with latest downloads and a Donate button in the upper right corner of each page
- http://code.opencv.org is for developers
- http://docs.opencv.org is the user manual, and
- http://code.opencv.org/projects/opencv/wiki/2012 has development-meeting notes
