https://bugs.kde.org/show_bug.cgi?id=503713
--- Comment #1 from Michael Miller <michael_mil...@msn.com> --- Hi glidias, There are many, many factors that will determine if using GPU acceleration will make facial recognition faster than CPU. While using the GPU for processing the face will make parts of the process faster, there is a cost for moving the information from main memory to GPU memory. On some systems moving the the image to/from the GPU might add more time than processing the image on the GPU saves. I recommend running some tests and picking the OpenCL setting that works best for you. I am working on updates to the face pipeline that will allow us to profile both CPU and GPU processing and automatically pick the best one, but this feature is very early in development. Also, we're exploring using OpenCV's Graph API (GAPI), which eliminates the OpenCL layer and moves the data closer to the GPU. Again, this feature is very early in development. Cheers, Mike -- You are receiving this mail because: You are watching all bug changes.
