Elador's profile - activity

Hmm what, first you describe QR codes, then you link to Eigen's QR decomposition? Very confusing what you're doing here. Actually the question is quite clear - QR decomposition relates to the matrix decomposition, and not to QR codes.

Just a quick question, how are you combing the PCA and LDA eigenvectors here, and why? gemm(pca.eigenvectors, leigen, 1.0, Mat(), 0.0, eigenvectors, GEMM_1_T);

As for your main question, what about using the distance to each cluster center as a score? If that doesn't work so well, try training SVMs?

Just two random comments: 1) You don't have to initialise trainData with the first image. Just start pushing back. Makes your code easier to understand if you loop from 0 everywhere. 2) std::string filename = "..." - you're copy-constructing a string from const char* here. Best do this: std::string filename("..."), or append 's' to make it a string literal (i.e. "..."s).

This doesn't work anymore in >=2.4.8, it always switches back to only building "Release;Debug".

Be careful, Matlabs reshape fills the data by filling the columns first, while OpenCVs reshape fills the rows first.

The following simple code fails with a

OpenCV Error: OpenGL API call (Can't load OpenGL extension [glGenerateMipmap]) i n IntGetProcAddress, file C:\opencv\2.4.7.2_prebuilt\opencv\sources\modules\core \src\gl_core_3_1.cpp, line 141

I built opencv with 2.4.7.2 with OpenGL enabled. The code below fails on Debug and Release mode, 64bit. I'm on Windows 7 - 64. I updated the graphics drivers (Intel GM45)... What is going wrong?

#include <string>

#include "opencv2/core/core.hpp"

#include "opencv2/highgui/highgui.hpp"

int main(int argc, char* argv[]) {

cv::Mat img = cv::Mat::zeros(640, 480, CV_8UC1);

cv::namedWindow("oglWin", cv::WINDOW_OPENGL);

cv::imshow("oglWin", img); // crash here

return 0;

}

The link to your fork is not working. User jlblancoc doesn't exist. Can you please update?

Is there a POSIT in the C++ API of OpenCV? I can only find the cvPOSIT of the old C API, and it's using all the old CvPoint3D32f and stuff, which is kind of ugly.

Also, what about when the old C API becomes deprecated in OCV 3.0?

I was able to remove my down-vote today. Didn't work 2 weeks ago.

Sorry, there's a misunderstanding here. The linked tutorial says "how to easily install OpenCv". But it does not install it, it shows how to make a _own_ project _using_ OpenCv, which is something different. For using OpenCv, of course CMake is not required (yet in my opinion still the easier solution than to add include paths and dozens of .lib's in every project manually). But judging from the initial post ("instaling", "pliz explain [...] tnx"), for that particular user, the linked tutorial indeed seems best. Sadly I cannot remove my down-vote, sorry for that.

Can you explain the downvote for suggesting ImageWatch, which is a 5-star vs2012 plugin and a perfect extension to the debugger for showing cv::Mat's?

Try ImageWatch, it's a great plugin for looking at cv::Mat's of all kinds (images and matrix-data).

Aside from that, I was never able to see the content of the Mat in text form in the debugger. There's a plus-sign, but it only shows some addresses etc, not the actual values inside the Mat.

Sorry, that's a VERY bad tutorial! You should never do that manually as specified in the tutorial. There is CMAKE. Follow installation instructions with cmake!

I have the same issue with VS2013. The error occurs with both OpenCV 2.4.6-stable and the git-master. The issue occurs in VS2013, but not in VS2012. How can we fix that for VS2013?

I was just thinking about the case when I have two windows and two instances MyClass, each having their draw-method bound to one of the windows. But I guess your proposed code also works in that case.

I don't think that's true, right? I mean why is there a CV_WINDOW_OPENGL and opengl_interop.hpp that specifically allows you to use opencv with opengl (and not viceversa)?

Thank you, I perfectly understand the problem now! I was now just wondering, could this be done with boost::bind/function? I've used bind before but only for very simple things, nothing like this.

Hi,

When I use

void on_opengl(void* userdata)
{
  // do stuff
}

int main(int argc, char *argv[])
{
  // stuff
  cv::setOpenGlDrawCallback("window", on_opengl, (void *)0);
  // stuff
}

it works fine. But when I try to move the callback-function inside a class:

Header:

namespace mylib {
class MyClass
{
public:
  MyClass();
private:
  void openGlDrawCallbackFunc(void* userdata);
};

CPP:

namespace mylib {

MyClass::MyClass()
{
  // stuff
  cv::setOpenGlDrawCallback("window", openGlDrawCallbackFunc, (void *)0); // errors on this line, on "openGlDrawCallbackFunc"
}

void MyClass::openGlDrawCallbackFunc(void* userdata)
{
  // stuff
}
}

it doesn't compile. The compiler gives the error

error C3867: 'mylib::MyClass::openGlDrawCallbackFunc': function call missing argument list; use '&mylib::MyClass::openGlDrawCallbackFunc' to create a pointer to member

Of course adding the "&" doesn't help. Intellisense shows a slightly different error,

Error: argument of type "void (mylib::MyClass::*)(void *userdata)" is incompatible with parameter of type "cv::OpenGlDrawCallback".

I included "opencv2/core/opengl_interop.hpp" and "opencv2/highgui/highgui.hpp", although the relevant stuff seems to be in highgui.hpp.

What is going wrong here and how can I do this?

Yea I generate my .sln also with cmake, but I always point the cmake of my own project to the opencv install dir, not the build-dir, as that's the proper way. But now I understand what you mean - I'll try that if 2.4.5 doesn't satisfy me for some reason, thank you :-)

Okay, thanks! That probably means for me I'll check out 2.4.5 and use that.

What do you mean with your suggestion "1."? I don't understand at all what you mean.

Concerning "2.", I never did that anyway! I always included the module headers, but as written above, their location has changed too.

I did not get any errors while compiling, what do you mean with am I "still" getting errors - maybe you confuse me with someone else?

Hiho,

a few weeks ago, I pulled opencv from git and compiled using cmake & vs2012. On my laptop, after compiling and running the INSTALL target, I get a structure like

`C:\opencv\install\include\opencv2\<moduleName>\someHeader.hpp`, e.g.
`C:\opencv\install\include\opencv2\core\core.hpp`.

In my code, I could then just #include "opencv2/core/core.hpp" and it was working perfectly in windows, linux, with people who used the downloadable-precompiled opencv - so all the paths seem to be right.

When pulling opencv from git today on another computer, also with cmake and vs2012, after compiling and INSTALL'ing, I get one folder

`C:\opencv\install\include`

and all the headers are just in there. There are no module subfolders. All headers are just in the "root". There is still a opencv2 subfolder there but it contains only 2 files. Of course, now my project fails compiling, as it can't find the includes under #include "opencv2/core/core.hpp".

I compiled opencv so many times, always used the same steps and never had this problem. What could possible be gone wrong on this computer?

Thanks!