Setting CMakeLists.txt for OpenCV GPU and CUDA

Autonomous Machines Jetson & Embedded Systems Jetson TX2
1

I couldn’t find a simple example of CMakeLists to compile my very first program using Opencv GPU like this:

#include <opencv2/opencv.hpp>
#include <opencv2/gpu/gpu.hpp> 
using namespace cv; 
... 
Mat image = imread("file.png"); 
gpu::GpuMat image_gpu; 
image_gpu .upload(image); 
gpu::GpuMat result; 
gpu::threshold(image_gpu, result, 128, CV_THRESH_BINARY); 
result.download(image); 
imshow("WindowName", image); 
waitKey ();

Can anyone suggest me a CMakeLists file with/without CUDA… please
and if you’ve any interesting tutorials to start with opencv gpu (mostly for compilation cmake…etc) would be great

2

Are you trying to compile this on the TX2? If so, the gpu:: libraries will not work if you’re using Opencv4Tegra right now. There was a compilation mistake which makes the gpu:: libraries unusable. The next update should have this fixed.

If you’ve build opencv from source, then a CMakeLists.txt file with contents similar to this should work!

cmake_minimum_required(VERSION X)
project(project_name)
find_package(OpenCV REQUIRED)
include_directories(${OpenCV_INCLUDE_DIRS})
add_executable(project_name source_file.cpp)
target_link_libraries(project_name ${OpenCV_LIBS})

You can find some OpenCV w/ CUDA and OpenCV w/o CUDA code on my GitHub page (ex: https://github.com/ShreyasSkandan/nvidia-tx1-tx2/tree/master/opencv_stuff/optical-flow)

Hope that answers your questions.

3

Yes I’m using TX2 and yes Opencv4Tegra is the one installed… So do you say that I can’t use opencv gpu functions ? shoud I install another version of opencv from source?

I can compile all my codes using opencv cpu. Now I want to move to gpu so I started with simple code (posted above in my question) but it doesn’t work with my cmakelistes nor with yours.

I checked your GitHub but it doesn’t have any code with GPU all with opencv cpu in which I don’t have any issues…

4

You can find more information regarding the problems with the gpu:: functions and Opencv4Tegra here:

https://devtalk.nvidia.com/default/topic/1000106/?comment=5114203
https://devtalk.nvidia.com/default/topic/1000701/?comment=5176691

You are advised to build from source if you wish to use the gpu:: functions immediately,

[url]https://devtalk.nvidia.com/default/topic/1000106/jetson-tx2/opencv-convertto-failure/post/5133788/#5133788[/url]

Yes, my GitHub code lacks any gpu:: functions because I’d rather not build from source at this point and lose some of the SIMT optimizations that Opencv4Tegra has.

5

Okay thanks for the links… the issue was posted the first time on 03/18/2017 so it is months ago and we don’t know when it will be fixed. I prefer to build from source for now.

6

Do you think that the same CMakeLists you gave me or that I use normally for CPU should work with gpu:: library?

7

Yup, it should/does work. I’ve tried the gpu:: libraries with the older Opencv4Tegra on the TX1 and it works!

8

I uninstalled OpenCV4Tegra from TX2 and installed the new version of OpenCv from source and it works (the gpu lib).

9

Hai, Can i ask u something bro?
When i run comand “cmake .”
I get this error.
Please help me.

I use ubuntu 16.04, cuda-8.0, opencv3.1

CMake Error at /usr/share/cmake-3.5/Modules/FindPackageHandleStandardArgs.cmake:148 (message):
Could NOT find CUDA: Found unsuitable version “7.5”, but required is exact
version “8.0” (found /usr)
Call Stack (most recent call first):
/usr/share/cmake-3.5/Modules/FindPackageHandleStandardArgs.cmake:386 (_FPHSA_FAILURE_MESSAGE)
/usr/share/cmake-3.5/Modules/FindCUDA.cmake:949 (find_package_handle_standard_args)
/usr/local/share/OpenCV/OpenCVConfig.cmake:48 (find_package)
/usr/local/share/OpenCV/OpenCVConfig.cmake:291 (find_host_package)
CMakeLists.txt:3 (find_package)

– Configuring incomplete, errors occurred!
See also “/home/panjaitan/Documents/HoughLine/CMakeFiles/CMakeOutput.log”.
See also “/home/panjaitan/Documents/HoughLine/CMakeFiles/CMakeError.log”.