Proof of concept that a Graphical Computational Unit can be used for real-time optimizations in visual systems.
Using CUDA in a consumer-grade Graphics Processing Unit to reach 1,000 speed up in an optimization problem
This work is protected by a Non-Disclosure Agreement. But here is what we can share about our exciting project.
A product is being developed by Cognizant Mobility that requires real-time image processing. One of the image correction algorithms is computationally intensive and it took 10 seconds to run in a consumer level CPU. The challenge was to perform the same optimization in the range of mili-seconds but implementation in a GPU was a challenge.
Using parallel computation algorithms and CUDA, the challenge was fully met (a 1,000 speed up) within weeks.
It was such a privilidge to work with such talented people at Cognizant Mobility.