mcExchange's profile - activity

Use docker installed OpenCV as library Is it possible to install OpenCV in a library and use it in my normal system (Ubu

Use docker installed OpenCV as library Is it possible to install OpenCV in a library and use it in my normal system (Ubu

true, answer works, my bad

What would be the same solution for opencv in python?

I guess it's kind of fine thanks! I still get some error while compiling, but that might be due to some missing linked library: error: undefined reference to 'cv::ml::SVM::create()'

I'm not sure entirely whether it's 3.0 or 3.1 I'm dealing with, since I'm doing surgery in the code of somebody else, but it's clearly not 2.4.x.

I'm new to opencv 3.0. I cannot figure out how to use a SVM object in a custom class. Can someone give me a toy example how to do it?

Here is an example for opencv 2.4.x:

// Class Definition    
class myClass{
    public:
        myClass();

    private:
        CvSVM svm;
};

// Constructor
myClass::myClass()
{
    myClass::svm.load("Some/hard/coded/path/to/a/SVM/File");
}

I guess the class definition should look similar to this:

class myClass {
    private:
    static cv::Ptr<cv::ml::SVM> svm;
};

But my Constructor keeps on failing:

myClass::myClass(){
    myClass::svm->load("Some/hard/coded/path/to/a/SVM/File");
}

with the following error message:

error: no matching function for call to 'cv::ml::SVM::load(const char [120])'

This question might apply to other objects as well.

I need to deconstruct a CvSVM object because I fear the risk of memory leakage. I know that in opencv Mat objects can be explicitly destroyed using the myMat.release() method. However I'm unsure how to do it for other objects like for example CvSVM.

Can I just call

CvSVM SVM;
delete SVM;

or is there another way?

I couldn't find any deconstructor method in the source code of CvSVM.

I need to run 3000 instances of kmeans and therefore would like to parllelize the code. However as soon as I'm starting to call kmeans from within a boost::thread group the results of kmeans look bad and are not random anymore. I guess inside a boost::threadGroup the random number generator just doesn't work anymore. Does anyone have a good suggestion how to solve this issue? Maybe there is even a better way to parralelize the execution of kmeans? Any help would be appreciated :)

:D Oh I didn't realize. Ok. Thanks for your answer

Could it be that the formula for computing the chi² distance between histograms as pointed out here is wrong?

If one divides the squared difference of (h1 and h2) by h1 the distance is not symmetrical anymore which is not logical.

Here is a paper that defines the chi square distance between histograms as:

1/2 * sum over i of (h1_i - h2_i)^2 / (h1_i + h2_i)

which is symmetric and makes more sense as a distance measure

Hi

In the wiki I found the following description of a function:

C++: Vec2d EM::predict(InputArray sample, OutputArray probs=noArray())

what does it mean exactly? I.e.

a) What exact type can "InputArray" be? A cv::Mat? or also a std::vector or also a simple C-style array?

b) What does the value noArray() specify? Is this a function? Or does it simply mean that I don't have to pass this argument because OpenCV can choose a default value here?