Vision APIs typically expect developers to specify how to solve a problem. Therefore state-of-the-art technologies are generally under-used due to either low awareness or time constraints. Under our what-oriented methodology for vision processing, the Open Vision Library (OpenVL) requires application developers to specify what they want the result to be, leaving the compiler/run-time implementations to determine how to do it.

Presented in Windsor, Ontario, May 2008 at the Canadian Conference on Intelligent Systems.