search

mdegans / nano_build_opencv

Build OpenCV on Nvidia Jetson Nano
MIT License
504 stars 199 forks source link

In jtop still cuda support No #73

Open mertmirac13 opened 1 year ago

mertmirac13 commented 1 year ago

I am a newbie and im using jetson nano for the first time and my problem is after compiling and installing cuda in jtop still cuda support says No. Is there a thing i should do after installation?

lweingart commented 1 year ago

Hi, I guess my issue is somewhat related to your. I have a Jetson AGX Orin that has opencv 4.5.4 installed by default, without cuda support. I used this repo on my other Jetson devices and it worked perfectly fine. This time, the build and compilation went well, the script was executed without errors.

However, when I import cv2, it's still the old version that is used, I don't know why. Additionaly, I cannot find a new cv2 directory (used to be created in /usr/local/lib) I find a lot of libopencv* lib files there, created by this install script, but I don't know how to use the newly installed opencv version.

Anyone could help ? Thank you very much in advance

Cheers

lweingart commented 1 year ago

Hi again,

I was actually able to solve my issue by applying the modifications suggested here: https://github.com/mdegans/nano_build_opencv/issues/67 As a summary: add 8.6 to the -D CUDA_ARCH_BIN option modify -D CUDNN_VERSION with 8.6 value and add this option: -D PYTHON3_PACKAGES_PATH=/usr/lib/python3/dist-packages

The last option, however, will have this newly installed opencv overwrite the default one, which in my case was just perfectly fine with me. Bye

mertmirac13 commented 1 year ago

Hi, my issue was same as yours but i solved it flashing my jetson nano and build cuda again. Thanks for your answers.

16 Nis 2023 Pzr 19:18 tarihinde Laurent Weingart @.***> şunu yazdı:

Hi again,

I was actually able to solve my issue by applying the modifications suggested here: #67 https://github.com/mdegans/nano_build_opencv/issues/67 As a summary: add 8.6 to the -D CUDA_ARCH_BIN option modify -D CUDNN_VERSION with 8.6 value and add this option: -D PYTHON3_PACKAGES_PATH=/usr/lib/python3/dist-packages

The last option, however, will have this newly installed opencv overwrite the default one, which in my case was just perfectly fine with me. Bye

— Reply to this email directly, view it on GitHub https://github.com/mdegans/nano_build_opencv/issues/73#issuecomment-1510424220, or unsubscribe https://github.com/notifications/unsubscribe-auth/AYVFNGZQ5RPXTYKYWWZSVPDXBQLVHANCNFSM6AAAAAAWJIHOA4 . You are receiving this because you authored the thread.Message ID: @.***>