Hi all,
Using ROS Melodic, OpenCV 3.4.6, and Leopard Imaging Cameras (IMX577’s). Our problem is that our pipeline purely uses the CPU and not the GPU. We are launching using Gstreamer and the following (not even running 4k). Can anyone give us some guidance on how to get better performance by using this awesome GPU?
nvarguscamerasrc sensor-id=0 ! video/x-raw(memory:NVMM), width=(int)1920, height=(int)1080,format=(string)NV12, framerate=(fraction)30/1 ! nvvidconv ! video/x-raw, format=(string)BGRx ! videoconvert ! video/x-raw, format=(string)BGR ! appsink
DaneLLL
December 30, 2020, 2:02am
4
Hi,
It is expected. Please check explanation in the post:
Hi,
It is limitation of hardware VIC engine. You have to use videoconvert plugin in the usecase.
We have DS4.0.1 for deep learning usecases. You may also check it and may be able to apply to your usease.
For optimal solution, we suggest run pure gstreamer pipeline or DeepStream SDK.
That is really bad news, because the cameras are a complete waste of money if I have to run them at such a low resolution to make up for the CPU.
What is meant by “pure gstreamer pipeline”? As in without OpenCV? Could you give me an example of what the format would look like please?
Hi,
You can refer to examples in gstreamer user guide .
If you need CUDA filters in OenCV, you may refer to