Jetson nano using GPU

Hey, everyone!
I wrote a code in python language using the OpenCV library and dnn for detecting costume objects.
I want the dnn and the OpenCV code to run on GPU.
i have Jetson nano Nvidia.
I have some questions:
1- I don’t understand this point: if I run my code on my Jetson nano, is it run ONLY on CPU?
2- I want to run on GPU , I read a lot about install CUDA, how to do that? every guide gives my errors …
3- the right thing to do is to connect the nano to the host pc and run the code on the CPU’s pc and some parts from the code will run on GPU?
4- i saw this: How to use OpenCV’s “dnn” module with NVIDIA GPUs, CUDA, and cuDNN <- is this what I need?
5- i saw numba library in python - nano Nvidia works with that?
I’m almost 3 days try to install/run my code on GPU without success …pls help me …

Default version of OpenCV installed with JetPack does not have CUDA/GPU-Support. You need to build&install OpenCV with CUDA support.
There are several methods/scripts for this floating around. I have success with this one: https://github.com/mdegans/nano_build_opencv

1 Like

i run the script ./build_opencv.sh 4.3.0
and i got failed in the building …
i attached the cmake log output and cmake error log:
CMakeOutput.log (74.5 KB) CMakeError.log (36.1 KB)

i glad if you can help me solve this… :)

Sorry, right now I don’t have time to look deeper into log files…

Which JetPack version is your Jetson running?
In case you are on JetPack 4.4 you need to build OpenCV 4.4.0 - see note in the nano_build_opencv repository…