How To Install and Build OpenCV C++ with NVIDIA CUDA GPU in Visual Studio Code

[@NicolaiAI]

Join My AI Career Program 👉 www.nicolai-nielsen.com/aicareer Enroll in My School and Technical Courses 👉 www.nicos-school.com

[@xToTaLBoReDoMx]

I love you for doing this!

[@NicolaiAI]

Thanks for watching! Really appreciate it

[@yaghiyahbrenner8902]

wow you rock simple to the point.

[@NicolaiAI]

Thanks for watching! Hope it could help u

[@cuongmanh368]

@@NicolaiAI hi nic, I have this error after compiling the project: OpenCV(4.9.0) Error: No CUDA support (The library is compiled without CUDA support) in throw_no_cuda, file C:\build\master_winpack-build-win64-vc17\opencv\modules\core\include\opencv2/core/private.cuda.hpp, line 106 OpenCV: terminate handler is called! Do you know how to fix it? When i config file on cmake, it already detected cuda and cudnn in the output.

[@danielginn411]

Really helpful thank you!

[@tajbaanook1597]

Please do a detailed project on cancer cell modeling and simulation. Thanks

[@joedavid8662]

If we were to generalize the instructions for an opencv installation without Cuda, what parameters in the video must be ignored and would there be any additional parameters?

[@NicolaiAI]

I can create a video where i just built OpenCV for cpu as well. Do u want to use it for something specific? If u just want a general installion u can just download the prebuilt OpenCV library

[@BdyPhrk]

Why can't I do run>start debugging or run>start without debugging it only works with ctrl+f5?

[@x-mesh8747]

when i am configure for second time its giving a no " modules has been found " error and i am copy paste the error message to chat gpt and it says you have to download codec how could i fix this

[@user-lz1rc6gk6r]

Hi, I have a question. While i was following the 'CMake Configuration' step, I couldn't find the 'OPENCV_DNN_CUDA'. But, I already installed cuDNN. So, i don't know what is wrong. Could you tell me how the 'OPENCV_DNN_CUDA' is appeared? Thanks.

[@x-mesh8747]

you have to add entry in cmake OPENCV_DNN_CUDA after that the problem has to fixed

[@neilsorkin7970]

Good tutorial but can you please figure out your mic. The clicking / volume drops are really driving me crazy.

[@NicolaiAI]

Thank you for letting me know, it was my video editor that caused the problem and is now fixed!

[@thehulk0111]

new video please with opencv 2022 because 2019 is not supported anymore and i cant even download it

[@Random-nf7qb]

Tried bulding 4.8.0 twice, and both times I'm missing the "install" folder. Any ideas? I have ticked the opencv_world in cmake. And the install folder was set.

[@Random-nf7qb]

@_ani_5065 uninstall everything from nvidia Then install cuda 12,1. Then build again.

[@Random-nf7qb]

@_ani_5065 well, that's exactly the issue. cuda 12,2 is not compatible. Uninstalling everything from nvidia and reinstalling 12,1 fixed it for me.

[@Random-nf7qb]

@_ani_5065 not sure about the cudnn tho, but for me it works with 8,9,3. If it doesn't for you, try 8.9.2.

[@prttyflvck0]

@_ani_5065have u got any solution?

[@Random-nf7qb]

@_ani_5065 did it work?

[@isaacberrios6463]

Thanks! This video was very helpful, but I have a question about fast math. I looked it up and seemed to only find horror stories of imprecise float roundoff. Will fast math be enabled by default? Or will we have to set a compiler flag to enable it?

[@anhphan8643]

thanks for this tutorial. i have a problem with the beginning. Even tried to install cuda toolkit 11.3, 11.2, 11.1 or 10.2, 10.1 but non of this worked, it kept warn "compute_50, sm_50 are deprecate" and i dont know how to fix this. Can you help me, many thanks.

[@NicolaiAI]

Never seen this error before. Have u tried to update or reinstall driver and cuda? Maybe some versions doesn't not match. Seems really weird if u can't even download Cuda from the Nvidia steps on their website

[@anhphan8643]

@@NicolaiAI Error MSB3721 The command ""C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v11.4\bin vcc.exe" -gencode=arch=compute_52,code=\"sm_52,compute_52\" --use-local-env -ccbin "C:\Program Files (x86)\Microsoft Visual Studio\2019\Community\VC\Tools\MSVC\14.29.30133\bin\HostX86\x64" -x cu -I"C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v11.4\include" -I"C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v11.4\include" -G --keep-dir x64\Debug -maxrregcount=0 --machine 64 --compile -cudart static -g -DWIN32 -DWIN64 -D_DEBUG -D_CONSOLE -D_MBCS -Xcompiler "/EHsc /W3 /nologo /Od /Fdx64\Debug\vc142.pdb /FS /Zi /RTC1 /MDd " -o x64\Debug\kernel.cu.obj "C:\Users\Administrator\source epos\CudaRuntime7\kernel.cu"" exited with code 1. C:\Program Files (x86)\Microsoft Visual Studio\2019\Community\MSBuild\Microsoft\VC\v160\BuildCustomizations\CUDA File: 11.4.targets Line: 785 I tried google but can't fix this.

[@anhphan8643]

i change to my output and got this: nvcc fatal : Could not set up the environment for Microsoft Visual Studio using 'C:/Program Files (x86)/Microsoft Visual Studio/2017/Community/VC/Tools/MSVC/14.16.27023/bin/HostX86/x64/../../../../../../../VC/Auxiliary/Build/vcvars64.bat'

[@NicolaiAI]

Do u have ms visual studio installed?

[@anhphan8643]

@@NicolaiAI yes, i tried both vs2019 and vs2017 (just install again this moring) but have not differrent

[@dev___2283]

With all respect sir i'm having a trouble at configuring in cmake it shows and error "Configuring incomplete,Error Occured!".....Help me sir what should i do or what should i look for!?

[@253_r.asidharth8]

Is the method same for a jetson xavier agx machine?

[@NicolaiAI]

Yeah u just have to make sure u choose the correct gpu architecture at the end of the configuration

[@OUKSEE]

Great video, thanks for this! I am having troubles running the program where a "No CUDA support" error appears. I assume the issue stems from my current opencv environmental paths. I watched your other CMake video and setup the environmental paths and such but for the new opencv downloads, would we need to use new environmental paths to target new lib and bin? If so, where are the corresponding lib and bin paths of the opencv-4.5.2 downloads?

[@OUKSEE]

would moving everything from your opencvTutorial folder in this video into the opencv folder made in your "How To Install OpenCV C++ and Set It Up in Visual Studio Code with CMake" video be the answer to my issue?

[@CosmJJ]

i have the same issue..

[@iiiFiinka]

If I plan to launch program on gtx 980 and gtx 1060 vide cards, can I set both values 5.2;6.1 in CUDAR_ARCH_BIN ? Thanks!

[@NicolaiAI]

Yes u can do that

[@iiiFiinka]

@@NicolaiAI thank you

[@oscarfuentes7866]

Help please. The first time I installed it without DNN support. So I tried to reinstall it by deleting my "build" folder and uninstalling Anaconda, to have everything clear. I tried to do the process again and now It doesn't work. Is there a proper way to uninstall completely the compiled OpenCV.

[@Digtal101]

Around 21:35 you mentioned/referenced towards another video you made about installing OpenCV with GPU support in the past. Can you leave a reference/link to the mentioned video? I've tried looking through your GitHub account and couldn't find it

[@davidaliaga4708]

the anaconda link is not there (although I have it already installed)

[@nurulistibudi9367]

hey thanks for the video, i have an geforce 940mx gpu hardware, it's said that it have cuda version 5.0 but i dont know and don't understand yet witch cudnn to choose for this version. thanks

[@kansairobot2015]

In min 22.34 you said "we have already included our directories from opencv in visua sudio code". When? and How did you do that?

[@NicolaiAI]

In the video a have about how to setup OpenCV with visual studio code 🙂

[@tayloreyler606]

@@NicolaiAI Oh! I didn't realize going through that video was a prerequisite for this video - it might be good to link to the other video at the beginning of this video in case there are others who didn't know there was a video we were supposed to follow before this one :)

[@NicolaiAI]

@@tayloreyler606 nope its not a a prerequisite but when I create the CMakelist file i just copy paste and say that i went over the steps in the other video i have for the interested

[@davidaliaga4708]

This method assumes that I have python installed in my PC, right? What if I have python not natively but with ananconda. RIght now the Cmake configuration is telling me that can not find python. What to do?

[@NicolaiAI]

Nope u can install it without python but then it will only work for c++ so yes the best idea is prob to install both at the same time. Anaconda should work fine and I have a video coming out about python in 5 minutes where i go over how to do it with Python aswell. There are some changes that need to be done so it doesn't cause errors. It says that the java, python2, python3 OpenCV module is unavailable right?

[@ROFLFISH]

Now I reached the part at 20:10. Unfortunately my VS looks completely different (but is 2019 Community Version 16.10.4) Instead of "File edit selection view go run terminal help" I have there: "File edit Git Debug Test Analyze Tools Extensions Window Help". And I am not able to bring up the command palette. The shortcut Ctrl + Shift + P doesn't work and under view is also not available.

[@NicolaiAI]

This is visual studio code and not visual studio 2019. That's in another video

[@ROFLFISH]

@@NicolaiAI If I type "cmake" in it says: "No matching commands"

[@NicolaiAI]

Thats because u don't have cmake as an environmental variable. U can tell to add it itself during installation of cmake, u can add the path to the cmake command file urself or type out the path instead of cmake in the terminal as u can't do now

[@ROFLFISH]

@@NicolaiAI I have: "C:\Program Files\CMake\bin" in my Path

[@NicolaiAI]

U should create a new variable "cmake" with the path

[@joepvand]

I am running into trouble with the dnn files not being made, despite checking all the required flags. Does anyone know why? cmake says cuDNN: YES

[@lukerobinson6720]

I just did all of this only realize the application I wanted to run requires 4.5.1 not 4.5.2 🙃is there a short way to redo all of this...

[@NicolaiAI]

Unfortunately not :(

[@lukerobinson6720]

​@@NicolaiAI rebuilding for 4.5.1 was the easy part. I just need to install cuDNN to an older version because it gives me an error like "[WARN] cuDNN is not running the same version that CUDART is reporting." I compiled opencv to run some random open source project that looks cool so that's the error it gives me when running said application. Will installing an older version of cuDNN on top of the newest work... might have to look into cleaning/purging the current install before installing the older version I need.

[@陳冠維-f7i]

I have never get this close to GPU computing before. But there is always something that stops me from reaching it. I was at the step of the Cmake command. there is no "install" folder in \opencv\build after the whole thing was built. edit: I used a really bad way of getting it to work(I installed the already compiled version without cuda support and swap out the opencv_world460.lib and .dll files. other headers are left un-touched.). I was able to get the cuda device info printed. But other functions like VideoCapture::read(), inRange(), GaussianBlur() became unrecognized external symbols, So do other functions.

[@prttyflvck0]

have u managed to do something with that? im facing the same issue

[@alexidesvignesnouvel1425]

Hi, I got some troubles to install CUDA to use it for a opencv project on QT. I use minGW as compiler and it appears that it can not work if I don't use Visual Studio as compiler (I got the error Cmake OpenCV is not able to find/configure CUDA SDK (required by WITH_CUDA)). Do you guys have a solution for that or is it really impossible to use CUDA and cuDNN with minGW?

[@davidaliaga4708]

How about WITH_TBB option? How can it be used?

[@NicolaiAI]

U can use that too if u want to. I have never worked with it before doe

[@kansairobot2015]

@@NicolaiAI In the past I have used that in linux. In linux it was just specifying -D USE_TBB but in windows it seems I had to install TBB somewhere and specify the path in CMake. So I didn't selected for now cause I want to have my windows working as quickly as possible

[@EDOARDOBONANNI-sd3dv]

Does this build allow running .cu files in addition to .cpp?

[@NicolaiAI]

Yeah u can do that but if u want to use them together u need to make sure u are working with the correct data type and structure in the kernels. If u are into python I actually have a course of how to use cuda together with opencv and have interfaces between those two. U can also implement and copy paste ur cuda kernels into cupy and use it in that way

[@akashprakash7511]

I am getting "nppi64.dll is missing" during program execution

[@NicolaiAI]

Hmm never seen that before

[@akashprakash7511]

@@NicolaiAI Thanks for the tutorial , just reinstalled and everything working fine. Thank you

[@NicolaiAI]

@@akashprakash7511 thanks for watching! Hope u can use it for some nice projects. My best guess would've been to do over again, great that it works now!

[@CosmJJ]

Please - need help! i get an error: Executing command: "C:\Program Files\CMake\bin\cmake.EXE" --build c:/Dev/OpenCV/VSCodeTest/build --config Debug --target ALL_BUILD -j 18 -- Build finished with exit code 0 OpenCV(4.5.3) Error: No CUDA support (The library is compiled without CUDA support) in throw_no_cuda, file C:\build\master_winpack-build-win64-vc15\opencv\modules\core\include\opencv2/core/private.cuda.hpp, line 106 OpenCV: terminate handler is called! Opencv works fine, but adding cuda - i get this error. N.B. I twice followed the video and reinstalled everything very carefully, but still get the same error. Any advice?

[@steevensonemile]

how can we use it on mac, please ?

[@NicolaiAI]

Only works with nvidia

[@rudrashisgorai9378]

Support ++

[@kansairobot2015]

Unfortunately I do this only on sundays but I got an update, perhaps you can help me. I have completed the video, the test works but when I try to build the simplest load (not even show) image example, it fails. I have investigated and turns out the problem is: 1. This installation only install the Release version of openCV right? 2. For some reason when I use cv::imread function in my code and even though I use ctrl F5 as in your video (that is supposed to be without debugging) it seems that it uses imread in Debug mode 3. Could you check where the exe file in the last part of your video is built? In my case it is build in the "Debug" directory, which is weird since it should be in the Release folder (that does not exist) Is there a way to build from visual studio code in Release mode? Should I reinstall everything with Debug version as well? Could you check that a simple load and display program works on your installation?

[@NicolaiAI]

U can choose down in the buttom that u want to build for release aswell. I'll try to reinstall with debug mode too and see it that fixes it

[@NicolaiAI]

I think it's because of the way it's build in vscode. In visual studio 2019 it works fine

[@kansairobot2015]

@@NicolaiAI That is right! very important. I checked down in the blue space where it says "CMake" and it said Debug. I clicked and chose Release and now it build for Release and it works well. Thanks!

[@NicolaiAI]

Great that it works now, sometimes things get a bit confusing with different text editors and IDEs especially when using CMake too since everything needs to be set to the same mode and etc. I'll try with debug mode too doe, just started the installation

[@venusshah4476]

I am not able to get any resulsts of install folder and the command we are typing in windows terminal

[@NicolaiAI]

Have u changed the path to your build folder?

[@venusshah4476]

No it is C:/my_path/build @@NicolaiAI

[@venusshah4476]

shoud I need to add the path in system or uservariable?

[@NicolaiAI]

No u should replace it with the path to the build folder u have created

[@venusshah4476]

also not ablee to find opencv GPU folder

[@jochenderycke9211]

I am not able to install ( 19:00 ). Do you know what this could mean. I get al lot of failed lines

[@jochenderycke9211]

I was trying to install while having the latest cuda version. But that did not work. I installed the cuda version used in this video and that worked :)

[@unamattina6023]

we installed opencv on previous video. does this step (3.15) is necessary?

[@unamattina6023]

up^+

[@screenwz]

'cmake' is not recognized as an internal or external command

[@syedabdulsalam4659]

make sure cmake is installed and in the path

[@screenwz]

@@syedabdulsalam4659 10 months

[@dirozx]

@@screenwz Too bad get real.

[@karishmathumu5257]

Hallo, I am trying this OPENCV with CUDa GPU installation on Jetson Orin. I am having this error: error: (-216:No CUDA support) The library is compiled without CUDA support in function 'throw_no_cuda' despite, enabling :opencv_dnn_cuda=ON, #include etc while configuring and testing the pkg-config --modversion opencv4 '4.5.4' Could you please help me solve my issue. Thank you

[@cuongmanh368]

hi, i have the same error, was yur problem solved?

[@karishmathumu5257]

@@cuongmanh368 yes, it was enabling OPENCV_WORLD while we enable the parameters on CMAKE-GUI.

[@cuongmanh368]

@@karishmathumu5257 tks, i also solved the problem already

[@ROFLFISH]

I don't know man. I wanted to use GPU in openCV with Python. So I started to watch this tutorial. But is says I should do this here first, to have some of the steps. No, this tutorial tells me to read tutorial x on cudnn and so forth. Maybe I am to dumb. But I think tutorials should not work this way. Just make a video in which EVERY step is done. I guess you lose 80% of the audience in the process of hopping from one video to another or a written guide.

[@NicolaiAI]

That is not correct, if i had to show the steps with Nvidia as well the video would be too long. And that's actually the only thing u have to do urself and no point in doing it when Nvidia perfectly tells u step by step what to do. And they keep the documentation up to date. Would u rather trust Nvidia or a KZbinr to edit and change ur GPU files on ur computer? And have the up to date documentation for the specific version u would use?

[@NicolaiAI]

I show all the steps in setting it up and doing the configurations but i don't show u how to install windows on ur computer too, that's a prerequisite as well.

[@ROFLFISH]

@@NicolaiAI I am just telling you that as beginner you want your instructor to go all the way with you. It is a lengthy installation process and some guys like me never used cmake, VS or Cuda before. So we/I need a more in detail explanation. Right now I reached the 20 minute mark and my computer is working, so I am optimistic. One more example in which I think more explanation would have been better: The command "cmake --build..." only works if cmake.exe is added to the Path. Just tell this. But nice that you are reading the comments.

[@NicolaiAI]

I'm answering all comments. When i mention the cmake installation i make it clear that u should check of the add cmake to path during installation. In 3 years time who will trust a KZbinr to modify their GPU files rather than just use to 3 lines that cuda provides. I even have to go in and check the versions as well. If u expect to install and use GPUs by just watching a video and not caring about what versions work together and so on, then u are going to have a really hard time. I'm not loosing 80% of my viewers because i tell them to use Nvidia documentation and check their GPU version rather than telling them to use my exact settings even doe I'm using other hardware and versions than them. That would just be dangerous and waste of time for them since it won't work. If u are new to everything and don't want to read documentation, i would definitely not recommend starting of with using GPUs and OpenCV if u lose focus and don't want to do it because u need 2 tabs open. Then it will be a nightmare going forward. The steps I'm doing will not mess with ur computer at all and if u just follow those steps and take 2 min to read 3 lines, then u will have it up and running in 1 hour + installation time. There u go. I'll only lose 80% of the viewers if a make a video that's General and will work in 1,3 and 5 years

[@kansairobot2015]

I finished installing and trying the test. But it is failing to find opencv. I suppose that in CMakeLists.txt we need to specify the path in find_package(OpenCV)?? Still not being able to find OpenCV. Why does it work in your case? I got CMake Error at CMakeLists.txt (find_package) By not providing FindOpenCV.cmake in CMAKE_MODULE_PATH this project has askec CMake to find a package configuration file provided by OpenCV,

[@NicolaiAI]

Do u have the path to the installation in the path environment variables? See the path in the description

[@mohamedaboraya4780]

@@NicolaiAI I did that and still getting error in find_package( OpenCV REQUIRED ) line

[@NicolaiAI]

@@mohamedaboraya4780 do u have the opencv_world file and all the other lib files? If u have configured it like in the video u only need to add the line from the description to the environment variables and then it is able to find the package

[@mohamedaboraya4780]

@@NicolaiAI I did everything in the video step by step except downloading anaconda and visual studio I just want to work with visual studio now and I have the same version. I configure BUILD_opencv_world just like what you did using CMake. Also added the path to the path environment variables. Any other idea what can cause this problem? 😅 Maybe there something I need to download and you explain in another video? To be more clear and specific this is the error:- "CMake Error at CMakeLists.txt:7 (find_package):By not providing \"FindOpenCV.cmake\" in CMAKE_MODULE_PATH this project has asked CMake to find a package configuration file provided by \"OpenCV\", but CMake did not find one CMake

[@NicolaiAI]

If u go into the build folder -> install and then there should be 2 files "opencvconfig" and "opencvconfig-version". Do u have those two? Those are the configurations for OpenCV with CMake. The first file is what's found when u do find_package(OpenCV required)

[@sabinkrshrestha9536]

MSBUILD : error MSB1009: Project file does not exist. Switch: INSTALL.vcxproj It throws this error while compiling in cmd, need help.

[@NicolaiAI]

Did u copy the command from the description? U can also check for the file in the build folder. If it's not there then its not configured correctly

[@hongyongzhang2301]

I compiled successfully when running github.com/opencv/opencv/blob/master/samples/gpu/video_reader.cpp What do I do if I add the with_opengl option at compile time，Error: The function/feature is not implemented (The called functionality is disabled for current build or platform) in throw_no_cuda, file E:\opencvgpu\opencv-4.5.2\modules\core\include\opencv2/core/private.cuda.hpp, line 112

[@NicolaiAI]

What do you mean at compile time? If u want to use opengl too u will need to choose it during the CMake configuration

[@hongyongzhang2301]

@@NicolaiAI Yes, the video_reader.cpp contains CV :: NamedWindow ("GPU", CV :: Window_OpenGL);So I need to check OpenGL in CMAKE. After I solve this problem, I get the following error. What should I do Error: The function/feature is not implemented (The called functionality is disabled for current build or platform) in throw_no_cuda, file E:\opencvgpu\opencv-4.5.2\modules\core\include\opencv2/core/private.cuda.hpp, line 112

[@NicolaiAI]

@@hongyongzhang2301 it certainly looks like the versions dont work together

[@hongyongzhang2301]

@@NicolaiAI What does that mean? I don't quite understand. If it can compile successfully, why can't it work together，Cuda10.1 cudnn7.6.5，What should I do, install a new CUDA? Which one? Have you tested it，Thank you very much for your reply

[@NicolaiAI]

I have not tried it with OpenGL so u will need to look at the versions and see if they support it. Seems weird that u can compile doe