CUDA and OpenCV

Robotics & Edge Computing Jetson & Embedded Systems Jetson Orin Nano
,
1

I have a Jetson Orin Nano 4GB developer kit, with the following features:
JetPack: 6.0
Ubuntu: 22.04
CUDA Arch BIN: 8.7
OpenCV: 4.5.4
Python: 3.10.12
Ultralytics: 8.3.26
L4T: 36.3.0
Torch: 2.3.0
Torchvision: 0.18.0
Numpy: 1.24.4
I attach an image; is it necessary that this mode “YES” the CUDA option to be able to detect objects in real time with YOLOv8? If so, how should I enable the option.

image
image866×458 65.6 KB

3

Hi,

The log shows your OpenCV doesn’t build with CUDA support.
(The default OpenCV in JetPack doesn’t have GPU support)

In general, YOLO from the Ultralytics team uses TensorRT for inference so OpenCV with CUDA should be optional.
But if you want to get the OpenCV with CUDA support, please build it from the source with the below script:

Thanks.

4

Hi, I followed the steps in the following page: Install OpenCV on Jetson Nano - Q-engineering, and it worked, I installed the OpenCV 4.10-dev version with CUDA “YES”, but GStreamer does not work, I guess it is essential to detect objects with Jetson Orin 4GB and with the IMX219-77 camera, if so how can I solve, thanks friend.

5

Hi,

GStreamer is another plugin so you will need to build OpenCV with GStreamer enabled.
For example, it should look like this:

cmake ... -D WITH_GSTREAMER=ON ..

Thanks.

8

This topic was automatically closed 14 days after the last reply. New replies are no longer allowed.