I’ll be concise.
GOAL: Apply DNN vision algorithms to a video data stream generated by a powerful, realistic simulator. All of this on a LINUX desktop.
MEANS: Since this is not implemented in the DRIVEWORKS API, develop this functionality (real time interface between a video flow and DNN object detector).
HOW: Use the SimpleCamera C++ class as starting point. Since, If I understood well, this class kinda simulate a camera video stream already.
As anybody a better idea ?
Better starting point ?
Objection to this idea (because unfeasible) ?