can visionworks improve video capture N resize speed?

I want to improve my webcam/video frame read speed.
I think nvx API also needs a frame from CV. is there a way to improve the reading speed using nvx API?
I guess opencv already would use multi threading inside to improve FPS.

Hi,

OpenCV by default use CPU-based FFmpeg as camera decoder.
It’s recommended to use GPU-accelerated framework to get better performance.

For USB-camera, please use GStreamer based pipeline.
Ex. VisionWorks, MMAPI, OpenCV with GStreamer support.

Thanks.