Currently, frame capture is done using the OpenCV decode function (both in cscore and libcamera backends) which is software only. At the minimum, Rockchip MPP and OMX based backends should be implemented to use hardware acceleration for frame decode.
Currently, frame capture is done using the OpenCV decode function (both in cscore and libcamera backends) which is software only. At the minimum, Rockchip MPP and OMX based backends should be implemented to use hardware acceleration for frame decode.