Reading CSI camera input directly to GPU memory

[Honey_Patouceul]

Capturing from nvarguscamerasrc with gstreamer, you would have the buffers into NVMM memory ready for GPU processing. If your processing doesn’t change resolution, you may use nvivafilter that can perform CUDA processing on NVMM buffers (RGBA or NV12, the latter may have stride constraint).
As an example of using opencv/cuda from nvivafilter, see:

Nvcompositor plugs alpha is not work and alpha plugs no work

nvivafilter may work as alpha plugin works, but nvcompositor would not be able to process it. Attached a nvivafilter: nviva-alpha-test.tgz (4.4 KB) just doing thresholding from hue to alpha. tar zxvf nviva-alpha-test.tgz cd alpha-test # You would edit Makefile and tune for your case (opencv install path...) make # in my case I used opencv-4.6.0 installed into /usr/local/opencv-4.6.0, so I first added (this may not be required for your case if you have a standard opencv intall, just for r…