marcob88's profile - activity

Hi, i've a questio (and if it is possible). I trying to check the difference between two compressed image of different quality. I try to check with

        for(int i=0; i < decompress_image_prev.rows; i++)
        {
            for(int j=0; j < decompress_image_prev.cols; j++)
            {
                if(image_to_send.at<uchar>(i,j) != decompress_image.at<uchar>(i,j) ||                   decompress_image.at<uchar>(i,j) != decompress_image_prev.at<uchar>(i,j))
                {
                    image_to_send.at<uchar>(i,j) = decompress_image.at<uchar>(i,j);
                    count++;                    
                }               
            }
        }

where decompress_image and decompress_image_prev are respect the actual and previous decompress image. But the final result is different from the last decompress image (quality = 100) and i've only the left part of image equal. Where is the problem?

Hi all,

i have a question (i suppose that it's stupid but i've difficult). I'm try to compile with cmake (not cmake-gui!!!) opencv. The problem is that i need to change the build folder. I give you an example:

I have the source code and the cmakelist on directory /home/xxx/opencv/

I would like to execute cmake and put the build into /home/yyy/lib_opencv/ and subsequently call make install....

With cmake-gui i don't have a problem. But i have to use the cmake command shell.

Can you help me? I hope that i have explained clearly the problem

Thanks

the problem is in Mat src = cv::imread(path, 0). I have a violation access memory for vcruntime140.dll lib

Hi,

i have a problem with opencv compiled for Visual Studio 2017.

I have a problem with this code

    for (String path : files)
    {
        Mat src = cv::imread(path, 0);

        if (!src.data)
        {
            cout << "Error" << endl;
            cout << "Path: " + path_considered << endl;
        }
     }

The errors is a violation access memory with vcruntime140.dll

Can you help me?

Thanks a lot

But if every mat have different size? For example i have a mat with 7 rows (number frame of video) and 10 cols ( number of words) but another mat can have different size...how can train in this situation? is possible?

Hi, I've a question with training an SVM multiclass with bag of word model. I have 50 video of training and each video is represented by a matrix of words.

Now i have a vector<mat> that contains all training data. How can training svm with this data structure?

I think to use pca but i'm not sure if is the exactly way.

Thanks

I'm try to use HOG People Detector for obtain Region of Interest. I have this depth image and the istruction is:

    HOGDescriptor hog;
hog.setSVMDetector(HOGDescriptor::getDefaultPeopleDetector());


vector<Rect> found, found_filtered;
hog.detectMultiScale(falseColorsMap, found, 0, Size(16, 16), Size(32, 32), 1.05, 2);
size_t i, j;
for (i = 0; i<found.size(); i++)
{
    Rect r = found[i];
    for (j = 0; j<found.size(); j++)
        if (j != i && (r & found[j]) == r)
            break;
    if (j == found.size())
        found_filtered.push_back(r);
}

for (i = 0; i<found_filtered.size(); i++)
{
    Rect r = found_filtered[i];
    r.x += cvRound(r.width*0.1);
    r.width = cvRound(r.width*0.8);
    r.y += cvRound(r.height*0.07);
    r.height = cvRound(r.height*0.8);
    rectangle(img, r.tl(), r.br(), Scalar(0, 255, 0), 3);
}

namedWindow("Display window", WINDOW_AUTOSIZE);// Create a window for display.
imshow("Display window", falseColorsMap);                   // Show our image inside it.

waitKey(0);                                          // Wait for a keystroke in the window
return 0;

how i can detect roi?

I can detect ROI in depth image on opencv 3.1? Can you help?

Hi, i try to develop a multiclass (n>2) classifier with svm. I saw that opencv implement one vs all for this problem. But i have a some problem with the parameter? How can choose best parameter for my classifier?

I develop a classifier fusion project, i use two SVM classifier then classify the same image and i need to develop a majority weighted voting process. How i can calculate weight from result of both classifier?

Hi, I have a problem with CamShift Tracking, At each iteration the region grows, reaching quickly the image size, and thus losing the person who is tracking. Where is the problem? This is my code

int main()
{
    string Directory = "C:\\Users\\Marco\\Desktop\\Scenario_1"; //directory dei file
    string StartFile = "gray-%06d.jpg";
    VideoCapture cap(Directory+"\\"+StartFile);
    int frameRef = 122;
    int countFrame = 1;
    Rect first(316, 94, 179, 380);
    Mat frame; Mat hsv;
    Mat mask, hue, roi_hist;
    float hranges[] = { 0, 180 };
    float sranges[] = { 0, 256 };
    const float* phranges = hranges;
    int hsize = 16;
    int hbins = 30, sbins = 32;
    //int histSize[] = { hbins, sbins };
    int histSize[] = { 180 };
    const float* ranges[] = { hranges };
    int channels[] = {1};
    TermCriteria term_crit(TermCriteria::EPS | TermCriteria::COUNT, 10, 1);
    Mat sec;
    Mat dest;
    bool isFirst = false;
    for (;;)
    {
        cap >> frame;

        if (frame.empty()) break;

        if (countFrame == frameRef)
        { 
            Mat firstFrame=frame(first);
            cvtColor(firstFrame, hsv, COLOR_BGR2HSV);
            inRange(hsv, Scalar(0, 60., 32.),Scalar(180., 255., 255.), mask);
            calcHist(&hsv, 1, channels, mask, roi_hist, 1, histSize, &phranges);
            //calcHist(&hsv, 1, channels, mask, roi_hist,2, histSize,ranges,true,false);
            //normalize(roi_hist, roi_hist, 0, 255, NORM_MINMAX);
            isFirst = true;
        }
        else if (isFirst)
        {
            cvtColor(frame, hsv, COLOR_BGR2HSV);
            calcBackProject(&hsv,1, channels, roi_hist,dest, &phranges);

            //calcBackProject(hsv, channels, roi_hist, dest, &phranges, 1.0);
            RotatedRect r=CamShift(dest, first, term_crit);


            //ellipse(frame, r, Scalar(0, 255, 0),3,LINE_AA);
            rectangle(frame, first, Scalar(0, 255, 0));
            namedWindow("ORIGINAL", WINDOW_AUTOSIZE);
            imshow("ORIGINAL", frame);

        }

        waitKey(2);
        countFrame++;
    }

}

Hi, I have a dataset where is a depth and rgb images taken by Kinect. And i have the information about bounding box on Rgb image, but not for depth, and this two sensor are not calibrate. I have also the intrinsic parameter, now there is a technique to obtain depth values of bounding box (x,y,width, height) from the information that I have?

Hi, i'm try to compress with webp codec but when i try to modify all image's pixel, the image will be changed only for left part. But with JPEG or other codec my program works well. Anyone have had this problem?

I have an image 640x480 and i need to extract a submatrix with this roi: x="216" y="107" width="423" height="255". I try to this:

    Mat frame = imread(path + fileName);
        vector<KeyPoint> keyPoint = vector<KeyPoint>();
        if (!frame.empty())
        {
            //prendo il punto del frame dell'evento interessato
            int width = itc->width;
            int height = itc->height;
            int x = itc->x;
            int y = itc->y;

            Mat bbox = Mat(frame,Rect(itc->x, itc->y, itc->x + itc->width, itc->y + itc->height));

but i have this exception :

Assertion Failed: 0 <= roi.x && 0 <= roi.width && roi.x + roi.width <= m.cols && 0 <= roi.y && 0 <= roi.height && roi.y + roi.height <= m.rows

where is the error? can you help me?

thanks, with this version i resolved the problem!!!