During May 23-24, there was a workshop called 'Edge Computing Using New
Commodity Architectures' at the University of North Carolina, Chapel Hill. As
the title says, the workshop is to review and discuss new computing
architectures such as GPUs.
My talk is basically the summary of the workshop, so I'll briefly introduce three new commodity architectures: Multi-core CPUs, The Cell Processor, and GPUs. Especially, I'll focus more on the GPU architecture because I have successfully implemented the GPU-based tensor voting which is about 30 times faster than the CPU-based implementation, and believe that your current framework also can be run fast with GPUs if it satisfies some requirements.