【Deepsort】C++版本Deepsort编译(依赖opencv,eigen3)

下载源码

https://github.com/shaoshengsong/DeepSORT

安装onnxruntime

安装方法参考博客

安装Eigen3

当谈及线性代数计算库时，Eigen3是一个强大而受欢迎的选择。Eigen3是一个C++模板库，提供了许多用于线性代数运算的功能，如向量、矩阵、矩阵运算、线性方程组求解等。以下是Eigen3的一些主要特点和功能：

高性能：Eigen3使用了优化的算法和技术，具有出色的运行速度和高效的内存使用。它通过使用表达式模板技术来在编译时生成高效的计算代码，避免了一些运行时的开销。

简洁易用：Eigen3提供了直观且易于使用的API，使得进行线性代数计算变得简单和直观。它的API设计简洁明了，提供了丰富的运算符重载和友好的语法，使得代码编写更加简单和可读。

广泛的功能：Eigen3支持多种线性代数运算，包括向量和矩阵的基本运算、线性方程组的求解、特征值和特征向量的计算、矩阵分解（如LU分解、QR分解等）、广义特征值问题的求解等。它还提供了一些高级功能，如支持稀疏矩阵和动态大小矩阵等。

跨平台支持：Eigen3是一个跨平台的库，可以在各种操作系统和编译器上使用。它使用标准的C++语法和特性，并且不依赖于任何特定的硬件或操作系统。

开源免费：Eigen3是一个开源库，遵循MPL2.0协议，可以免费使用和修改。它的源代码可在GitHub上获得，方便用户进行定制和扩展。
Eigen3已被广泛应用于科学计算、机器学习、计算机图形学等领域。无论是在学术研究还是实际应用中，Eigen3都是一个强大而可靠的线性代数库。

sudo apt-get install libeigen3-dev

编译opencv

编译命令：下载opencv4的源码包后解压，并下载contrib扩展包，开启dnn
读取视频需要安装ffmpeg:

sudo apt-get install build-essential
sudo apt-get install cmake git libgtk2.0-dev pkg-config libavcodec-dev libavformat-dev libswscale-dev
sudo apt-get install python-dev python-numpy libtbb2 libtbb-dev libjpeg-dev libpng-dev libtiff-dev libjasper-dev libdc1394-22-dev

对于libjasper-dev安装失败：

apt-get install software-properties-common
add-apt-repository "deb http://security.ubuntu.com/ubuntu xenial-security main"
apt update
apt-get install libjasper-dev

sudo apt-get install ffmpeg

获取ffmpeg的相关安装目录：

pkg-config --cflags libavformat

编译：

cmake    -DCMAKE_BUILD_TYPE=Release -DOPENCV_EXTRA_MODULES_PATH=/home/opencv_contrib-4.7.0/opencv_contrib-4.7.0/modules -DWITH_OPENCL=OFF -DBUILD_DOCS=OFF -DBUILD_EXAMPLES=OFF -DBUILD_WITH_DEBUG_INFO=OFF -DBUILD_TESTS=OFF -DWITH_1394=OFF -DWITH_CUDA=OFF -DWITH_CUBLAS=OFF -DWITH_CUFFT=OFF -DWITH_OPENCLAMDBLAS=OFF -DWITH_OPENCLAMDFFT=OFF -DINSTALL_C_EXAMPLES=OFF -DINSTALL_PYTHON_EXAMPLES=OFF -DINSTALL_TO_MANGLED_PATHS=OFF -DBUILD_ANDROID_EXAMPLES=OFF -DBUILD_opencv_python=OFF -DBUILD_opencv_python_bindings_generator=OFF -DBUILD_opencv_apps=OFF -DBUILD_opencv_calib3d=OFF -DBUILD_opencv_features2d=OFF -DBUILD_opencv_flann=OFF -DBUILD_opencv_java_bindings_generator=OFF -DBUILD_opencv_js=OFF -DBUILD_opencv_ml=OFF -DBUILD_opencv_objdetect=OFF -DBUILD_opencv_photo=OFF -DBUILD_opencv_python3=OFF -DBUILD_opencv_python_tests=OFF -DBUILD_opencv_shape=OFF -DBUILD_opencv_stitching=OFF -DBUILD_opencv_superres=OFF -DBUILD_opencv_ts=OFF -DBUILD_opencv_videostab=OFF -DBUILD_opencv_world=ON -DBUILD_opencv_dnn=ON -D WITH_FFMPEG=ON -D WITH_TIFF=OFF -D BUILD_TIFF=OFF -DWITH_FFMPEG=ON  -DFFMPEG_LIBRARIES=/usr/local/lib/ -D FFMPEG_INCLUDE_DIRS=/usr/local/include/ ..make install

更改deepsort对应的cpp文件：

#include <fstream>
#include <sstream>
#include <opencv2/imgproc.hpp>
#include <opencv2/opencv.hpp>
#include <opencv2/dnn.hpp>
#include "YOLOv5Detector.h"#include "FeatureTensor.h"
#include "BYTETracker.h" //bytetrack
#include "tracker.h"//deepsort
//Deep SORT parameterconst int nn_budget=100;
const float max_cosine_distance=0.2;void get_detections(DETECTBOX box,float confidence,DETECTIONS& d)
{DETECTION_ROW tmpRow;tmpRow.tlwh = box;//DETECTBOX(x, y, w, h);tmpRow.confidence = confidence;d.push_back(tmpRow);
}void test_deepsort(cv::Mat& frame, std::vector<detect_result>& results,tracker& mytracker)
{std::vector<detect_result> objects;DETECTIONS detections;for (detect_result dr : results){//cv::putText(frame, classes[dr.classId], cv::Point(dr.box.tl().x+10, dr.box.tl().y - 10), cv::FONT_HERSHEY_SIMPLEX, .8, cv::Scalar(0, 255, 0));if(dr.classId == 0) //person{objects.push_back(dr);cv::rectangle(frame, dr.box, cv::Scalar(255, 0, 0), 2);get_detections(DETECTBOX(dr.box.x, dr.box.y,dr.box.width,  dr.box.height),dr.confidence,  detections);}}std::cout<<"begin track"<<std::endl;if(FeatureTensor::getInstance()->getRectsFeature(frame, detections)){std::cout << "get feature succeed!"<<std::endl;mytracker.predict();mytracker.update(detections);std::vector<RESULT_DATA> result;for(Track& track : mytracker.tracks) {if(!track.is_confirmed() || track.time_since_update > 1) continue;result.push_back(std::make_pair(track.track_id, track.to_tlwh()));}for(unsigned int k = 0; k < detections.size(); k++){DETECTBOX tmpbox = detections[k].tlwh;cv::Rect rect(tmpbox(0), tmpbox(1), tmpbox(2), tmpbox(3));cv::rectangle(frame, rect, cv::Scalar(0,0,255), 4);// cvScalar的储存顺序是B-G-R，CV_RGB的储存顺序是R-G-Bfor(unsigned int k = 0; k < result.size(); k++){DETECTBOX tmp = result[k].second;cv::Rect rect = cv::Rect(tmp(0), tmp(1), tmp(2), tmp(3));rectangle(frame, rect, cv::Scalar(255, 255, 0), 2);std::string label = cv::format("%d", result[k].first);cv::putText(frame, label, cv::Point(rect.x, rect.y), cv::FONT_HERSHEY_SIMPLEX, 0.8, cv::Scalar(255, 255, 0), 2);}}}std::cout<<"end track"<<std::endl;
}void test_bytetrack(cv::Mat& frame, std::vector<detect_result>& results,BYTETracker& tracker)
{std::vector<detect_result> objects;for (detect_result dr : results){if(dr.classId == 0) //person{objects.push_back(dr);}}std::vector<STrack> output_stracks = tracker.update(objects);for (unsigned long i = 0; i < output_stracks.size(); i++){std::vector<float> tlwh = output_stracks[i].tlwh;bool vertical = tlwh[2] / tlwh[3] > 1.6;if (tlwh[2] * tlwh[3] > 20 && !vertical){cv::Scalar s = tracker.get_color(output_stracks[i].track_id);cv::putText(frame, cv::format("%d", output_stracks[i].track_id), cv::Point(tlwh[0], tlwh[1] - 5),0, 0.6, cv::Scalar(0, 0, 255), 2, cv::LINE_AA);cv::rectangle(frame, cv::Rect(tlwh[0], tlwh[1], tlwh[2], tlwh[3]), s, 2);}}}
int main(int argc, char *argv[])
{//deepsorttracker mytracker(max_cosine_distance, nn_budget);//bytetrackint fps=20;BYTETracker bytetracker(fps, 30);//-----------------------------------------------------------------------// 加载类别名称std::vector<std::string> classes;std::string file="./coco_80_labels_list.txt";std::ifstream ifs(file);if (!ifs.is_open())CV_Error(cv::Error::StsError, "File " + file + " not found");std::string line;while (std::getline(ifs, line)){classes.push_back(line);}//-----------------------------------------------------------------------std::cout<<"classes:"<<classes.size();std::shared_ptr<YOLOv5Detector> detector(new YOLOv5Detector());detector->init("/home/DeepSORT-master/DeepSORT-master/build/yolov5x.onnx");std::cout<<"begin read video"<<std::endl;const std::string source = "/home/DeepSORT-master/DeepSORT-master/build/test.mp4";cv::VideoCapture capture(source);if (!capture.isOpened()) {printf("could not read this video file...\n");return -1;}std::cout<<"end read video"<<std::endl;std::vector<detect_result> results;int num_frames = 0;cv::VideoWriter video("out.avi",cv::VideoWriter::fourcc('M','J','P','G'),10, cv::Size(1920,1080));while (true){cv::Mat frame;if (!capture.read(frame)) // if not success, break loop{std::cout<<"\n Cannot read the video file. please check your video.\n";break;}num_frames ++;//Second/Millisecond/Microsecond  秒s/毫秒ms/微秒usauto start = std::chrono::system_clock::now();detector->detect(frame, results);auto end = std::chrono::system_clock::now();auto detect_time =std::chrono::duration_cast<std::chrono::milliseconds>(end - start).count();//msstd::cout<<classes.size()<<":"<<results.size()<<":"<<num_frames<<std::endl;//test_deepsort(frame, results,mytracker);test_bytetrack(frame, results,bytetracker);//cv::imshow("YOLOv5-6.x", frame);video.write(frame);if(cv::waitKey(30) == 27) // Wait for 'esc' key press to exit{break;}results.clear();}capture.release();video.release();cv::destroyAllWindows();}