Mostafa Hagog, Architect at Intel, presents the "Developing High-Performance, Energy-Efficient Vision Solutions Using Open-Standard Libraries and APIs" tutorial at the May 2015 Embedded Vision Summit.
Connected cameras, hand-held devices, embedded systems, and cloud services enable a world of intelligence everywhere. But to realize the potential of computer vision technology, developers face tough challenges. On the one hand, complex products must be brought to market quickly. On the other hand, computer vision applications demand both high performance and extreme energy-efficiency.
Open standard libraries and APIs like OpenCL, OpenVX, and OpenCV offer the promise of faster and easier development. But can these libraries and APIs also enable high-performance, energy-efficient implementations?
In this presentation, Mostafa highlights which aspects of vision software development tend to be similar across different types of platforms (e.g., mobile vs. cloud) and which tend to be different—and why. Using real-world examples, he explores how open standard libraries and APIs can be used to speed application development while minimizing loss of performance and energy efficiency—or in some cases, even improving them.