Pedro Batista's profile - activity

I am attempting to calculate sums over rectangular areas in images using the Viola & Jones trick with the integral image.

Im using c++ and the cv::Mat class.

Here is what I am doing:

.
.
.


    integral(Img,M);

    CvRect region;

    region.x=0; 
    region.y=0; 
    region.width=Img.cols;
    region.height=Img.rows; //calculating sum on entire image


      double regionSum;

      int tl= IntegralImage.at<double>((region.y),(region.x));
      int tr= IntegralImage.at<double>((region.y),(region.x+region.width+1));
      int bl= IntegralImage.at<double>((region.y+region.height+1),(region.x));
      int br= IntegralImage.at<double>((region.y+region.height+1),(region.x+region.width+1));

      regionSum = br-bl-tr+tl;

      cout<<"tl: "<<tl<<" tr: "<<tr<<" bl: "<<bl<<" br: "<<br<<" sum: "<<regionSum<<endl;

For some images those parameters output zero, for others there is a result but that doesn't really make sense, so obviously I am doing something wrong-

Thanks in advance.

I've created an openCV application for human detection on images.

I run my algorithm on the same image over different scales, and when detections are made, at the end I have information about the bounding box position and at which scale it was taken from. Then I want to transform that rectangle to the original scale, given that position and size will vary.

I've wrapped my head around this and I've gotten nowhere. This should be rather simple, but at the moment I am clueless.

Help anyone?

This method will work with multiple hands in the same image if each one is a contour on its own. Once you detect fingert

@Raki, if you end up trying Steven's suggestion I would be curious to know the results!

I gave you the direction and now it is your job to do the work. Use cv2.findContours() to calculate the countour coordin

Maybe try taking a different aproach, try to compute all distances between the hand centroid and each point in the conto

Maybe try taking a different aproach, try to compute all distances between the hand centroid and each point in the conto

Maybe try taking a different aproach, try to compute all distances between the hand centroid and each point in the conto

Maybe try taking a different aproach, try to compute all distances between the hand centroid and each point in the conto

Maybe try taking a different aproach, try to compute all distances between the hand centroid and each point in the conto

Maybe try taking a different aproach, try to compute all distances between the hand centroid and each point in the conto

Maybe try taking a different aproach, try to compute all distances between the hand centroid and each point in the conto

Maybe try taking a different aproach, try to compute all distances between the hand centroid and each point in the conto

CNNs are trully remarkable. It is an established fact that they are solving problems that years of previous academic and

I realize that I may sound like a guy that is not able to adapt to a change in the field, but that is not the case. I us

I realize that I sound like a guy that is not able to adapt to a change in the field, but that is not the case. I use wh

CNNs are trully remarkable. It is an established fact that they are solving problems that years of previous academic and

It will increase if your video is so small that the splitting function would become the bottleneck, which would make you

It will increase if your video is so small that the splitting function would become the bottleneck, in which case making

Why don't you split your video into multiple files in a preprocessing function, and then create an array of VideoCapture

I've had and solved this particular problem. Here is a description of what I did. Unfortunately the links to the images

You want a rotated rectangle containing all the contours in the vector, our just a bounding box enclosing all the contou

It seems that versions prior to 3.1.0 do not have all FileStorage functionality implemented. After version 3.2.0 this

It seems that versions prior to 3.0.0 do not have all FileStorage functionality implemented. After version 3.2.0 this

I don't think opencv has a standard way to calculate the magnitude of a given line, so, if you have the coordinates of e

An example image would help to give a more definitive answer, but yes, you can probably automate this process using Open

An example image would help to give a more definitive answer, but yes, you can probably automate this process using Open

An example image would help to give a more definitive answer, but yes, you can probably automate this process using Open

An example image would help to give a more definitive answer, but yes, you can probably automate this process using Open

An example image would help to give a more definitive answer, but yes, you can probably automate this process using Open

An example image would help to give a more definitive answer, but yes, you can probably automate this process using Open

An example image would help to give a more definitive answer, but yes, you can probably automate this process using Open

An image would help to give a more definitive answer, but yes, you can probably automate this process using OpenCV, but

I accidentelly misclicked into deleting a comment from this thread, sorry to the autor, please repost the comment.

If you have a problem, post a question with code :) That information is not enough to know what is wrong.

What opencv version are you using? I found out that only after 3.2.0 this was implemented in Python.

OpenCV expects a contour to be a list of points of connected pixels, in which the starting point is the top-most one (wi

OpenCV expects a contour to be a list of points of connected pixels, in which the starting point is the top-most one (wi

Generally plase give me some suggestions what is you impression about such a system. Depends on a lot of things.

It makes sense that reducing the sliding window stride will improve the detection accuracy, because a smaller stride wil

how to detect multiple cars in an image using opencv?? I'm using the following code, but i'm able to detect only one car

While developing some stuff that uses blob analysis , I realized my application was dropping frames. After evaluating what might be the cause, I singled out the findContours call as the culprit.

I created this small code snippet to understand what was going on,

#include <opencv2/core.hpp>
#include <opencv2/imgproc.hpp>
#include <opencv2/highgui.hpp>
int main()
{
    while ((char) cv::waitKey(30) != 'q')
    {
        cv::Size size = cv::Size(640, 480);

        cv::Mat testFindContours = cv::Mat::zeros(size, CV_8UC1);

        std::vector<std::vector<cv::Point> > contours;
        std::vector<cv::Vec4i> hierarchyObj;

        cv::findContours(testFindContours, contours, hierarchyObj, cv::RETR_EXTERNAL, cv::CHAIN_APPROX_SIMPLE);

        cv::imshow("Black window", testFindContours);
    }

    return 0;
}

Weirdly enough, just by using findContours on a 640x480 completely black image, my CPU spikes to ~300% usage (more than one CPU cores at full speed).

The number on the most right-hand side is the CPU usage % (from "top" in linux)

Even weirder, is when I reduced the Size to 320x240, now the CPU usage stayes below 2%

I started playing a bit with the size, and I have the impression that there is a specific size at which the CPU spikes up to 300%, but I didn't go far enough to check which size that is exactly, so I am not completely sure about this.

What am I missing here? Is it normal that findContours even consumes anything on a completely black image? Might this be a bug?

The CPU I am doing these tests in is a intel core i7 3.4 ghz and opencv version 3.2.0. I'm compiling this code in Debug configuration.

Best regards.