Problems of copying gstreamer buffer into numpy array

[LoveNvidia]

Hi guys,
For the gstreamer buffer(memory:nvvm) we can convert buffer to numpy array, This occurred a critical problem: copy from GPU buffer to CPU buffer,
I want to know, this problem only is for discrete GPUs or involved also for jetson devices? If it’s for both, In the jetson devices the GPU and CPU memory is shared, why this copying is occurred? I want to know those buffers (GPU buffer and CPU buffer) are independent from shared memory?

[DaneLLL]

Hi,
There are several posts with similar questions:

How to pass GstBuffer to inference plugin in custom gstreamer plugin

Hi all, @mdegans, @miguel.taylor I want to create a custom gstreamer buffer and added to deepstream. My goal is to create inference plugin gstreamer like face recognition plugin and add to deepstream pipeline. This work has two Question: I want to use python in do_transform_ip() method: 1- How to pass gst-buffer to inferecne model? I don’t want to convert buffer to numpy array. 2- Crop the buffer with given coordinates(metadata) that comes from before element and then feed to model.

Write custom gstreamer plugin in python and use tin deepstream

Hi, Q1- Suppose I create a pipeline and the last pipeline produce with caps : video/x-raw(memory:NVMM) and I want to create a custom plugin and perform some processing on nvidia jetson devices. This need to do processing in GPU buffer? Is it possible to convert buffers to numpy? If it so, Is it optimal solution? Q2- I create a class that inherited from GstBase.BaseTransform and has this method: def do_transform_ip(self, buffer: Gst.Buffer) -> Gst.FlowReturn: // Do some processing on B…

This is known in python programming. We would suggest write a custom plugin in C instead of using numpy array.