frondeur's profile - activity

Even tried to compile opencv with CUDA_ARCH_BIN = 1.1 and PTX=1.1 but no vain, the problem still persists, the error now apears when initializing a gpumat e.g. prev_imgMatGPU=cv::gpu::GpuMat(image);

and error points to the line in the gpumat.cpp

void mallocPitch(void* devPtr, size_t step, size_t width, size_t height) const { cudaSafeCall( cudaMallocPitch(devPtr, step, width, height) ); }

I can see in the debugger that image is not empty, it has width and height, so i don't know what actually is troubling it.

Thanks, actually i didn't know when they moved it into separate lib. So when i figured it out the problem was solved.

Hi,

I am using opencv2.4.2 with visual studio 2010 build with cmake (CUDA enabled). The problem is i consistently get a crash when accessing gpu methods/initializations while being in debug mode. It works in release mode although that too needs to be run outside visual studio otherwise some crash occur. The exception that i receive is as follows:

First-chance exception at 0x76dbfbae in opencvGPUtest.exe: Microsoft C++ exception: cudaError_enum at memory location 0x00a3e258

The error is very consistent, and i have seen couple of examples over the internet with others (http://tech.groups.yahoo.com/group/OpenCV/message/77905). It is very difficult to work without debug support while working with large projects, so kindly help me fix this.

Thanks,

Actually the cuda arch variable as i understand is relative to the graphics card's compatibility so that is the reason i put it 1.1 (for NVIDIA quadro 140M). Anyway i have checked with 2.0 being set for both which comes as a default value too but the problem persists. The qt\bin path is in environemnt as i can access from the shell qmake.exe. Cuda version is 4.2 Although i have noticed that you have qt inside opencv as a dependency whereas i have downloded qt from the qt website and its version is 4.8.0 while yours seem to be 4.8.2

Thanks for the reply. I am using VS2010 for compilation and has version 14.

Microsoft (R) 32-bit C/C++ Optimizing Compiler Version 14.00.50727.762 for 80x86

I have seen such a problem elsewhere with others too so it might not be related to CUDA as i can run cuda opencv samples. It works in Release mode but fails in debug mode. In release mode i get another issue which is a crash at "GPUMat" definition. Although this also works fine outside visual studio in release mode and fails to work inside vs2010 or in debug.

Regards,

I have checked from the documentation about the type of FASTGPU return type but it is not listed there. So then i have tried printing from the code and it returns '5' which means 'CV32F' . They are keypoints so shoudn't it be possible to pass them to pyrLK even though they have slightly different type?

Hi,

I am trying to track feature using optical flow method. The features are corners extracted from fast as a gpu::GpuMat keypoints. The problem is that the call to "sparse" function fail with following error:

OpenCV Error: Assertion failed (prevPts.rows == 1 && prevPts.type() == CV32FC2) in unknown function, file H:\opencv\modules\gpu\src\pyrlk.cpp, line 114 CV error occured : C:\opencv\modules\gpu\src\pyrlk.cpp:114: error: (-215) prevPt s.rows == 1 && prevPts.type() == CV32FC2

I've checked that the previous keypoints that i am passing to it are non zero. Although i have two rows in gpu::GpuMat. Is that correct? Actually i tried the fast_GPU documentation but i couldn't understand "Location Row"/"Response Row". Even if i use only single row from it for passing to sparse, i still get the assertion. It is unclear how could i make the features points from fast to be compatible with optical flow function? Kindly help in this regard.

Best Regards,

I've figured it out. In new versions of opencv the SURF/SIFT are distributed in a separate library which needs to be initialized before creation of feature detector.

I've noticed one thing that changing the detector method to FAST makes the same code work. Does this mean that SURF is not being implemented in opencv anymore?

Hi,

feature detector fails when used in simple program in visual studio 2010. I am using opencv 2.4.2 and also checked on 2.4.1. Only thing that is being done is to create a feature detector and use that to detect keypoints in an image. I get unhandled exception crash pointing to a function named "detecImpl()" inside detectors.cpp (i.e. features2d\detectors.cpp line:65). This error is really stuck and has taken enormous amount of time so any help is really appreciated.

#include <iostream>
#include <opencv2/core/core.hpp>
#include "opencv2/highgui/highgui.hpp"
#include <opencv2/imgproc/imgproc.hpp>
#include <opencv2/features2d/features2d.hpp>

using namespace std;
using namespace cv;

int main(int argc, char* argv[])
{
cv::Ptr<cv::FeatureDetector> featureDetector;
    cv::Ptr<cv::DescriptorExtractor> descriptorExtractor;
    featureDetector = cv::FeatureDetector::create("SURF");
    descriptorExtractor = cv::DescriptorExtractor::create("SURF");
cv::Mat imageColor;
cv::Mat image = cv::imread("car1.jpg", 0);
    cv::cvtColor(image, imageColor, CV_GRAY2BGR);
    try{
        imshow("Test Image",imageColor);
        cv::waitKey(3000);
    }
    catch(cv::Exception exc)
    {
        cout << "CV error occured : " + exc.msg;
    }
std::vector<cv::KeyPoint> currentKeypoints;

try{
            featureDetector->detect(image,currentKeypoints);   //This line generates the error but no exception is caught ....
    }
    catch(cv::Exception exc)
    {
        cout << "CV error occured : " + exc.msg;
        return -1;
    }
}

Hi,

I've opencv 2.4.2 build with visual studio 2010 and CMAKE (qt support enabled). Whenever, i use a simple program opening a window/use imshow then i get unhandled exception crash and in the terminal window i get opencv error :

OpenCV error: Null pointer <null session="" handler=""> in unknown function, file, pathtoopencv\highgui\src\window_qt.cpp line 909

The same program with the same configuration settings work in 'Release' configuration. It only fails in Debug mode. I also tried to build the opencv with SSE/SSE2 disabled but still am getting the crash.

Regards,