I have two questions.
First, Is it possible to hardcode an OpenCV algorithm onto a GPU without needing a CPU/OS?

Secondly, if the answer to the first question was a NO... then what is the lowest CPU processing speed that would be necessary for me to process video (720p) from a GPU(such as a BCM3835) with a simple Blob+Color Detection algorithm with OpenCV and stream it with SSH?

If either of these sound a little to vague or like there are holes within the questions(i.e. more info needed), I can clarify more.

Thank you

 

I have two questions.
First, Is it possible to hardcode an OpenCV algorithm onto a GPU without needing a CPU/OS?

I don't see how.

 

You possibly could interface the standalone GPU over pci-e bus, for example from an fpga. However I don´t see why would you want to go through such trouble when you can use the PC, drivers and OS to ease your work.