Revision history [back]

Hello, This seems a bit trivial, but I'm not sure as to why execution isn't working how I'm expecting it to. I want to resize an image and then apply the Gaussian blur to that image, but only one is being done; whichever comes last.

Am I losing the data from the first operation as it goes to the second?

And how could I also do this is parallel with multiple GPUs if I wanted to use streams? If that's at all possible.

All help is appreciated Thank you!

Hello, This seems a bit trivial, but I'm not sure as to why execution isn't working how I'm expecting it to. I want to resize an image and then apply the Gaussian blur to that image, but only one is being done; whichever comes last.

Am I losing the data from the first operation as it goes to the second?

And how could I also do this is parallel with multiple GPUs if I wanted to use streams? If that's at all possible.

All help is appreciated Thank you!

Hello, This seems a bit trivial, but I'm not sure as to why execution isn't working how I'm expecting it to. I want to resize an image and then apply the Gaussian blur to that image, but only one is being done; whichever comes last.

Am I losing the data from the first operation as it goes to the second?

And how could I also do this is parallel with multiple GPUs if I wanted to use streams? If that's at all possible.

All help is appreciated Thank you!

Hello, This seems a bit trivial, but I'm not sure as to why execution isn't working how I'm expecting it to. I want to resize an image and then apply the Gaussian blur to that image, but only one is being done; whichever comes last.

Am I losing the data from the first operation as it goes to the second?

And how could I also do this is parallel with multiple GPUs if I wanted to use streams? If that's at all possible.

All help is appreciated Thank you!

Hello, This seems a bit trivial, but I'm not sure as to why execution isn't working how I'm expecting it to. I want to resize an image and then apply the Gaussian blur to that image, but only one is being done; whichever comes last.

Am I losing the data from the first operation as it goes to the second?

And how could I also do this is parallel with multiple GPUs if I wanted to use streams? If that's at all possible.

All help is appreciated Thank you!

void resizeWithGauss(std::string inputFile, std::string outputFile)
{
outFile){
Mat inputHost = imread(inputFile, CV_LOAD_IMAGE_COLOR);
 cuda::GpuMat inputDevice(inputHost);
 cuda::GpuMat outputDevice;
 const int ksize = 21;
 const int type = CV_64F;
 Timer timer;
 timer.Start();
 cuda::resize(inputDevice, outputDevice, Size(), 2.0, 2.0, CV_INTER_CUBIC);
 cv::Ptr<cuda::filter> cv::Ptr<cuda::Filter> gauss = cv::cuda::createGaussianFilter(inputDevice.type(), outputDevice.type(), Size(ksize, ksize), 6.0, 6.0);
 gauss->apply(inputDevice, outputDevice);
 timer.Stop();
 printf("OpenCV GPU code ran in: %f msecs. \n", timer.ElapsedTime());
 Mat outputHost;
 outputDevice.download(outputHost);
 imwrite(outputFile, outputHost);
 inputHost.release();
 outputDevice.release();
}
outputDevice.release();}