Currently, typical open source facial recognition algorithms are fast enough to provide results quickly, but not necessarily fast enough to process data in real-time, such as a video stream. In an effort to remedy this, an open source facial recognition algorithm should be redeveloped in either C++ or Java using CUDA libraries to utilize the parallelization capability of a GPU.
- Develop a system to read a video stream and apply a facial recognition algorithm leveraging the capabilities of a GPU to the video frames.
- Provide a list of people identified in the video and the time at which they were seen
This will allow one to process data for facial recognition in real-time.
Develop a facial recognition algorithm which leverages a GPU based off of the math of a pre-existing algorithm
This will require a CUDA capable GPU and an environment for development of CUDA code (all open-source software).
The solution developed will become intellectual property of Scitor, a SAIC company.
Point of Contact
Scitor, a SAIC Company