Hello everybody. I'd like to know if there is any example of TBB and Opencv. I'm hoping to run, let's say applying two filters to the same image, and each of the filters running in different cores? Is it possible?
Thank you.
Yes, you can do like this:
// A very raw example of using tbb's thread
#include <opencv2/core.hpp>
#include <opencv2/imgproc.hpp>
#include <opencv2/video/video.hpp>
#include <opencv2/highgui.hpp>
#include <tbb/tbb.h>
#include <iostream>
using namespace std;
using namespace cv;
using namespace tbb;
int main(int argc, char* argv[])
{
Mat im1 = imread(argv[1]);
Mat imGray;
if (im1.data == nullptr)
{
cout << "Error while reading file " << argv[1];
return 1;
}
imshow("Input image", im1);
cvtColor(im1, imGray, CV_RGB2GRAY);
Mat im3, im4;
tbb_thread th1([&imGray, &im3]() // in fact, you can do this with C++ thread
{
int windowSize = 5; // starting threshold value
int constant = 5; // starting constant value
adaptiveThreshold(imGray, im3, 255,
CV_ADAPTIVE_THRESH_MEAN_C, CV_THRESH_BINARY,
windowSize, constant);
});
tbb_thread th2([&imGray, &im4]()
{
cv::GaussianBlur(imGray, im4, cv::Size(3, 3), 5.0f);
});
th1.join();
th2.join();
imshow("Grayscale image", imGray);
imshow("Adaptive thresholding", im3);
imshow("Gaussian blur", im4);
cvWaitKey(0);
return 0;
}
As gaussianBlur use thread too I think you will have 2*getNumthreads() threads when you th2.join
It's speed is not important parameter (like sometimes in GUI where user interaction is important) that's not a problem.
AFAIK, all the statements after the join() call can deal with results obtained from previous threads (whose join() were called). So if you fed image (s) to one thread to process, after its join() finishes (join() statement in the program), you can be sure to cope with other images (or the ones produced from the thread).
What I look for is to create a loop with a listener function which will obtain a signal when some thread function have finished. I cannot imagine how it works. There should be some feeding function running in a loop in memory, which will create queue when it is needed or will pass the images from queue to the thread function. I just cannot imagine how to manage the feeding process. I will create question for this tomorrow because I just compiled OpenCV with TBB and I am completely new to this. So I first need to try your code and then I will ask.
You can use the TBB parallel pipeline with OpenCV to do just that.
I have a post about using it with OpenCV here - https://www.theimpossiblecode.com/blo...
TBB has advanced multi threading patterns and tools - some of them used internally by OpenCV (if built with it), and some can be used externally, regardless if OpenCV was built with TBB support.
Asked: 2015-10-07 17:14:14 -0600
Seen: 6,810 times
Last updated: Oct 08 '15
Parallel processing is used very offen in OpenCV particulary in filter processing. If you want to use TBB for filtering it's always used in different way. Image is divided in strip and multi-core is used for each strip.
I think If you want to use TBB for each image you have to disable TBB in Opencv and activate TBB in your program.
There is no example for TBB because TBB is encapsulated like OpenMP COncurrency cstripes and GCD.
(what is GCD and CStripes?)
There is an interesting article "The Foundations for Scalable Multi-core Software in Intel® Threading Building Blocks" http://citeseerx.ist.psu.edu/viewdoc/... with fibonacci number example similar as is in example code within TBB package. See page 54. It's long article but contains precious examples, explaining, diagrams and benchmarks. Link to LBerger's note: http://answers.opencv.org/question/15...