Build and install OpenCV from source with CUDA and cuDNN support

-------- IMPORTANT --------
Please add OPENCV_GENERATE_PKGCONFIG=1 flag when configuring to create the opencv.pc so other applications can find opencv.
If "nvcc not found" then create following soft links to local bin.
sudo ln -s /usr/local/cuda-11.3/bin/* /usr/local/bin
sudo ln -s /usr/local/cuda-11.3/nvvm/bin/* /usr/local/bin
-------- END --------
-------- IMPORTANT --------
cuDNN installation has changed since this video. Please refer the instruction in the documentation
docs.nvidia.co...
-------- END --------
CUDA instillation guide:
docs.nvidia.co...
cuDNN installation guide:
docs.nvidia.co...
OpenCV repositories:
github.com/ope...
github.com/ope...
CUDA wiki page:
en.wikipedia.o...
Follow me on:
Email: [e-mail]
GitHub: github.com/s1n7ax
LinkedIn: / srinesh-nisala
Facebook: / s1n7ax
Instagram: / s1n7ax

[@kunalk1234]

This was indeed very helpful. I was not able to build with cuda support but now I can. Thanks a ton !

[@s1n7ax]

Good if you only use cuda. cudnn installation has changed slightly since this video.

[@kunalk1234]

CUDNN installation was a big pain and I had to copy files. Cmake configuration detected cudnn as well. I think only DNN module uses cudnn which I don't plan to use as of now. Cuda accelerated libraries should work fine.

[@MrGopak]

Exact and thorough manual. Need to add Ubuntu 20.04 in the title for better indexing. Huge thanks!

[@YoussifAshrafEdwar]

you are the best explanation

[@tungphamkhachoang1625]

That was very helpful, thank you so much!

[@anhdang2077]

ông cài theo được ko ông

[@zakariaesaidi843]

Thankyou, it worked for me

[@arndtsteffens7831]

Very well done. Thank you

[@s1n7ax]

--------- IMPORTANT --------- cuDNN installation has changed since this video. Please refer the instruction in the documentation docs.nvidia.com/deeplearning/cudnn/install-guide/index.html#package-manager-ubuntu-install

[@simplesecomumcidadao7592]

9:52 When I paste the link in this section and press enter, it gives me an error and says no such command was found. how can I do it

[@Abhishek_Satasiya]

I have done all the things you did in the video. and all things worked out perfectly fine. but in the last step when I write import cv2 it is giving me ModuleNotFoundError: No module named 'cv2'

[@s1n7ax]

Someone else said it worked so you could be doing something wrong. Pay attention to errors. Easier to debug

[@recepalpsaygl8484]

did you solve the problem

[@guillermoalejandrobastianf7330]

I installed and didn't have errors. But when I run the final command "cv2.cuda.getCudaEnabledDeviceCount()" is returning me 0. Could be because I have CV2 installed using pip before?

[@karishmathumu5257]

This was very informative and great narrative. But i am having issues when i am doing the 2nd round of configuration. It says "Error in configuration process, project files maybe invalid". Could you please help me.

[@aliesatgungor]

Nice Video. Thanks

[@timewaster7101]

Hi @s1n7ax! Thank for your help but I face some trouble to install open CV. Could you answer my question If I use (base) - virtual environment created by anaconda. can I import cv2 in (base) to used in pytorch? and I installed Cuda and CuDNN but not like you so any difference between your install process and my. Sorry for my bad english and thank you so much for your help

[@s1n7ax]

--------- IMPORTANT --------- Please add OPENCV_GENERATE_PKGCONFIG=1 flag in cmake-gui so other applications can find opencv. If "nvcc not found" then create following soft links to local bin. sudo ln -s /usr/local/cuda-11.3/bin/* /usr/local/bin sudo ln -s /usr/local/cuda-11.3/nvvm/bin/* /usr/local/bin

[@mohamedkhalilallaoui4747]

Nice video thanks ! . I have a question i noticed that you imported cv2 globally what if i use virtualenv in a project , how can i use the opencv gpu on my venv ??

[@s1n7ax]

Change the include directory to virtual env's include dir

[@comingsoon578]

i was succeed built cv2 with cmake, installed it perfectly and cuda support also. but how to put this libs into a virtualenv

[@s1n7ax]

Try creating the virtual environment first. Use the virtual environment python and include paths. I think that should work.

[@comingsoon578]

@@s1n7ax thanks a lot sir, it works 👍

[@eddygonzalez6513]

Is it possible to install this with anaconda ?

[@s1n7ax]

Not with cuda enable i guess

[@eddygonzalez6513]

@@s1n7ax Could you verify that CUIDA is working from the vscode? is that in my case it does not take the interpreter /usr/bin/python3 it only uses /usr/bin/python and when changing it throws an error (Ubuntu 22.04). I would like the only interpreter that uses ubuntu to be that of anaconda3