我正在尝试学习CUDA中的编程。我创建了一个简单的应用程序，使用OpenCV在avi文件中进行对象检测，然后尝试使用CUDA加快检测速度。在这一点上，我正在努力正确地做它。因此，我的计划是使对象检测在不同的线程中运行以加快应用程序的速度。这是我到目前为止所做的基本代码：
#include <iostream>
#include <iomanip>
#include "opencv2/objdetect.hpp"
#include "opencv2/highgui.hpp"
#include "opencv2/imgproc.hpp"
#include "opencv2/cudaobjdetect.hpp"
#include "opencv2/cudaimgproc.hpp"
#include "opencv2/cudawarping.hpp"
#include "arcos.h"
#include "cuda.h"
#include "cuda_runtime.h"
using namespace std;
using namespace cv;
using namespace cv::cuda;
void help()
{
cout << "Usage: ./cascadeclassifier \n\t--cascade <cascade_file>\n\t(<image>|--video <video>|--camera <camera_id>)\n"
"Using OpenCV version " << CV_VERSION << endl << endl;
}
void convertAndResize(const Mat& src, Mat& gray, Mat& resized, double scale)
{
if (src.channels() == 3)
{
cv::cvtColor(src, gray, COLOR_BGR2GRAY);
}
else
{
gray = src;
}
Size sz(cvRound(gray.cols * scale), cvRound(gray.rows * scale));
if (scale != 1)
{
cv::resize(gray, resized, sz);
}
else
{
resized = gray;
}
}
void convertAndResize(const GpuMat& src, GpuMat& gray, GpuMat& resized, double scale)
{
if (src.channels() == 3)
{
cv::cuda::cvtColor(src, gray, COLOR_BGR2GRAY);
}
else
{
gray = src;
}
Size sz(cvRound(gray.cols * scale), cvRound(gray.rows * scale));
if (scale != 1)
{
cv::cuda::resize(gray, resized, sz);
}
else
{
resized = gray;
}
}
void matPrint(Mat &img, int lineOffsY, Scalar fontColor, const string &ss)
{
int fontFace = FONT_HERSHEY_DUPLEX;
double fontScale = 0.8;
int fontThickness = 2;
Size fontSize = cv::getTextSize("T[]", fontFace, fontScale, fontThickness, 0);
Point org;
org.x = 1;
org.y = 3 * fontSize.height * (lineOffsY + 1) / 2;
putText(img, ss, org, fontFace, fontScale, Scalar(0, 0, 0), 5 * fontThickness / 2, 16);
putText(img, ss, org, fontFace, fontScale, fontColor, fontThickness, 16);
}
void displayState(Mat &canvas, bool bHelp, bool bGpu, bool bLargestFace, bool bFilter, double fps)
{
Scalar fontColorRed = Scalar(255, 0, 0);
Scalar fontColorNV = Scalar(118, 185, 0);
ostringstream ss;
ss << "FPS = " << setprecision(1) << fixed << fps;
matPrint(canvas, 0, fontColorRed, ss.str());
ss.str("");
ss << "[" << canvas.cols << "x" << canvas.rows << "], " <<
(bGpu ? "GPU, " : "CPU, ") <<
(bLargestFace ? "One, " : "Multi, ") <<
(bFilter ? "Filter:ON" : "Filter:OFF");
matPrint(canvas, 1, fontColorRed, ss.str());
// by Anatoly. MacOS fix. ostringstream(const string&) is a private
// matPrint(canvas, 2, fontColorNV, ostringstream("Space - switch GPU / CPU"));
if (bHelp)
{
matPrint(canvas, 2, fontColorNV, "Space - switch GPU / CPU");
matPrint(canvas, 3, fontColorNV, "M - switch One / Multi");
matPrint(canvas, 4, fontColorNV, "F - toggle rectangles Filter");
matPrint(canvas, 5, fontColorNV, "H - toggle hotkeys help");
matPrint(canvas, 6, fontColorNV, "1/Q - increase/decrease scale");
matPrint(canvas, 7, fontColorNV, "R - rotate on/off");
}
else
{
matPrint(canvas, 2, fontColorNV, "H - toggle hotkeys help");
}
}
void elforgat(Mat& src, double fok)
{
Point2f center(src.cols / 2.0f, src.rows / 2.0f);
Mat rot = cv::getRotationMatrix2D(center, fok, 1.0);
Mat f;
cv::warpAffine(src, f, rot, src.size());
src = f;
}
void elforgat(GpuMat& src, double fok)
{
Point2f center(src.cols / 2.0f, src.rows / 2.0f);
Mat rot = cv::getRotationMatrix2D(center, fok, 1.0);
GpuMat f;
cuda::warpAffine(src, f, rot, src.size());
src = f;
}
__global__ void detectLyme(cv::Ptr<cv::cuda::CascadeClassifier> &cascade_gpu, bool findLargestObject, double scaleFactor, bool filterRects, cv::Size &minSize, cv::Size &maxSize, cv::cuda::GpuMat &resized_gpu, cv::cuda::GpuMat &facesBuf_gpu, std::vector<cv::Rect> &faces)
{
cascade_gpu->setFindLargestObject(findLargestObject);
cascade_gpu->setScaleFactor(1 / scaleFactor);
cascade_gpu->setMinNeighbors((filterRects || findLargestObject) ? 4 : 0);
cascade_gpu->setMinObjectSize(minSize);
cascade_gpu->setMaxObjectSize(maxSize);
cascade_gpu->detectMultiScale(resized_gpu, facesBuf_gpu);
cascade_gpu->convert(facesBuf_gpu, faces);
}
int main(int argc, const char *argv[])
{
VideoCapture capture;
Mat image;
string inputName;
bool isInputImage = false;
bool isInputVideo = true;
bool isInputCamera = false;
Ptr<cuda::CascadeClassifier> cascade_gpu;
cv::CascadeClassifier cascade_cpu;
Mat frame, frame_cpu, gray_cpu, resized_cpu, frameDisp;
vector<Rect> faces;
GpuMat frame_gpu, gray_gpu, resized_gpu, facesBuf_gpu;
bool useGPU = true;
double scaleFactor = 0.9;
bool findLargestObject = false;
bool filterRects = true;
bool forgass = false;
bool helpScreen = false, run = true;
double fok = 0;
TickMeter tm;
Size minSize = Size(30, 30);
Size maxSize = Size(34, 34);
double detectionTime, fps;
char key;
string cascadeName = "cascade.xml";
inputName = "D:\\cicc\\video\\085\\Teszt40\\Teszt_Teszt40_2018-10-09-17-09-25_free_1.avi";
setlocale(LC_ALL, "hun");
if (getCudaEnabledDeviceCount() == 0)
return cerr << "No GPU found or the library is compiled without CUDA support" << endl, -1;
//cuda::printShortCudaDeviceInfo(cv::cuda::getDevice());
cuda::printCudaDeviceInfo(cv::cuda::getDevice());
cascade_gpu = cuda::CascadeClassifier::create(cascadeName);
cout << "name: " << cascade_gpu->getDefaultName() << " size " << cascade_gpu->getClassifierSize() << endl;
if (!cascade_cpu.load(cascadeName))
return cerr << "ERROR: Could not load cascade classifier \"" << cascadeName << "\"" << endl, help(), -1;
if (isInputImage)
{
image = imread(inputName);
CV_Assert(!image.empty());
}
else if (isInputVideo)
{
capture.open(inputName);
CV_Assert(capture.isOpened());
}
else
{
capture.open(atoi(inputName.c_str()));
CV_Assert(capture.isOpened());
}
namedWindow("lajm-e", 1);
while (run)
{
if (isInputCamera || isInputVideo)
{
capture >> frame;
if (frame.empty()) run = false;
}
if (run) // end ?
{
tm.reset();
if (useGPU)
{
tm.start();
frame_gpu.upload(image.empty() ? frame : image);
if (forgass) elforgat(frame_gpu, fok);
convertAndResize(frame_gpu, gray_gpu, resized_gpu, scaleFactor);
//detectLyme(cascade_gpu, findLargestObject, scaleFactor, filterRects, minSize, maxSize, resized_gpu, facesBuf_gpu, faces);
detectLyme<<<1, 256>>>(cascade_gpu, findLargestObject, scaleFactor, filterRects, minSize, maxSize, resized_gpu, facesBuf_gpu, faces);
tm.stop();
resized_gpu.download(resized_cpu);
}
else
{
tm.start();
(image.empty() ? frame : image).copyTo(frame_cpu);
convertAndResize(frame_cpu, gray_cpu, resized_cpu, scaleFactor);
if (forgass) elforgat(resized_cpu, fok);
cascade_cpu.detectMultiScale(resized_cpu, faces, 1.05,
(filterRects || findLargestObject) ? 4 : 0,
(findLargestObject ? CASCADE_FIND_BIGGEST_OBJECT : 0)
| CASCADE_SCALE_IMAGE
, minSize, maxSize);
tm.stop();
}
if (forgass) fok++;
for (size_t i = 0; i < faces.size(); ++i)
rectangle(resized_cpu, faces[i], Scalar(255));
detectionTime = tm.getTimeMilli();
fps = 1000 / detectionTime;
//print detections to console
cout << setfill(' ') << setprecision(2);
cout << setw(6) << fixed << fps << " FPS, " << faces.size() << " det";
if ((filterRects || findLargestObject) && !faces.empty())
for (size_t i = 0; i < faces.size(); ++i)
cout << ", [" << setw(4) << faces[i].x
<< ", " << setw(4) << faces[i].y
<< ", " << setw(4) << faces[i].width
<< ", " << setw(4) << faces[i].height << "]";
cout << endl;
cv::cvtColor(resized_cpu, frameDisp, COLOR_GRAY2BGR);
displayState(frameDisp, helpScreen, useGPU, findLargestObject, filterRects, fps);
imshow("lajm-e", frameDisp);
key = (char)waitKey(1);
switch (key)
{
case 27:
run = false; break;
case ' ':
useGPU = !useGPU; break;
case 'm': case 'M':
findLargestObject = !findLargestObject; break;
case 'f': case 'F':
filterRects = !filterRects; break;
case '1':
scaleFactor *= 1.05; break;
case 'q': case 'Q':
scaleFactor /= 1.05; break;
case 'h': case 'H':
helpScreen = !helpScreen; break;
case 'r': case 'R':
forgass = !forgass; break;
}
} // if run
} // while (run)
cout << "File v�ge" << endl;
//if (key!=27) std::cin.get();
return 0;
}
有人可以帮助我如何优化代码以加快速度吗？