How to implement batch processing on Xavier ?

I plan to implement batch proessing on Xavier.

I try to use OpenCV and GStreamer as input/output interface.
And inference executed by tensorRT.

I had refered exmples of tensorrt
and understood the way to create the engine which could do batch processing.

I am facing how to implement I/O.
For example, is it possible to combine four inputs into one and make inferences?
I know that Deepstream has nvstreammux. How do I extract video data when dealing with this in C ++?

Hi,

Deepstream can support batch process and multiple input.

[source0]
enable=1
#Type - 1=CameraV4L2 2=URI 3=MultiURI 4=RTSP
<b>type=3</b>
uri=file://../../streams/sample_1080p_h264.mp4
num-sources=4
...

[primary-gie]
enable=1
<b>batch-size=4</b>
...

Thanks.

Thank you, AastaLLL.

I’ll try the way to use nvstreammux.