ubuntu clion从0开始搭建一个风格转换ONNX推理网络 opencv cuda::dnn::net

系统搭建

• 系统搭建

OpenCV的安装

cmake

sudo apt-get install cmake

其他环境以来

sudo apt-get install build-essential libgtk2.0-dev libavcodec-dev libavformat-dev libjpeg.dev libtiff5.dev libswscale-dev libjasper-dev  

• 不安装会报这个错误

OpenCV(4.6.0) /home/dell/下载/opencv-4.6.0/modules/highgui/src/window.cpp:1250: error: (-2:Unspecified error) The function is not implemented. Rebuild the library with Windows, GTK+ 2.x or Cocoa support. If you are on Ubuntu or Debian, install libgtk2.0-dev and pkg-config, then re-run cmake or configure script in function 'cvNamedWindow'

error

• E: 无法定位软件包 libjasper-dev

解决（换源重新安装）

• http://mirrors.ustc.edu.cn/help/ubuntu.html

sudo add-apt-repository "deb http://security.ubuntu.com/ubuntu xenial-security main"
sudo apt update

• 由于没有公钥,无法验证下列签名: NO_PUBKEY 40976EAF437D05B5 NO_PUBKEY 3B4FE6AC

sudo apt-key adv --recv-keys --keyserver keyserver.ubuntu.com 40976EAF437D05B5 3B4FE6ACC0B21F32

下载源码

https://github.com/opencv/opencv/releases
增强模块 https://github.com/opencv/opencv_contrib/tags

安装

• unzip opencv-4.6.0.zip
• unzip opencv_contrib-4.6.0.zip
• cd opencv-4.6.0/
• sudo mkdir build
• cd build
• sudo cmake -D CMAKE_BUILD_TYPE=Release -D CMAKE_INSTALL_PREFIX=/usr/local
  或 sudo cmake -D CMAKE_BUILD_TYPE=Release -D CMAKE_INSTALL_PREFIX=/usr/local -D OPENCV_EXTRA_MODULES_PATH= **/opencv_contrib-4.6.0/modules/ ..

error

• 如果报错 CMake Error: The source directory "/home/dell/下载/opencv-4.6.0/build" does not appear to contain CMakeLists.txt. 使用 sudo cmake /home/dell/下载/opencv-4.6.0/ -DCMAKE_BUILD_TYPE=Release -DCMAKE_INSTALL_PREFIX=/usr/local

• sudo make或 sudo make -j4 j为变异是使用的核心数量,这一步非常满
• sudo make install

-- Installing: /usr/local/share/opencv4/lbpcascades/lbpcascade_profileface.xml
-- Installing: /usr/local/share/opencv4/lbpcascades/lbpcascade_silverware.xml
-- Installing: /usr/local/bin/opencv_annotation
-- Set runtime path of "/usr/local/bin/opencv_annotation" to "/usr/local/lib"
-- Installing: /usr/local/bin/opencv_visualisation
-- Set runtime path of "/usr/local/bin/opencv_visualisation" to "/usr/local/lib"
-- Installing: /usr/local/bin/opencv_interactive-calibration
-- Set runtime path of "/usr/local/bin/opencv_interactive-calibration" to "/usr/local/lib"
-- Installing: /usr/local/bin/opencv_version
-- Set runtime path of "/usr/local/bin/opencv_version" to "/usr/local/lib"
-- Installing: /usr/local/bin/opencv_model_diagnostics
-- Set runtime path of "/usr/local/bin/opencv_model_diagnostics" to "/usr/local/lib"

环境变量

• sudo vim /etc/ld.so.conf.d/opencv.conf i /usr/local/lib + esc + :wq + enter
  
• sudo ldconfig
• sudo vim /etc/bash.bashrc

PKG_CONFIG_PATH=$PKG_CONFIG_PATH:/usr/local/lib/pkgconfig  
export PKG_CONFIG_PATH  

• sudo updatedb

或者按照这种方式配置

测试

cmake_minimum_required(VERSION 3.15)
project(untitled)
set(CMAKE_CXX_STANDARD 14)MESSAGE(STATUS "Project: untitled")               #打印相关消息消息
find_package(OpenCV REQUIRED)# 通过find_package引入非官方的库（该方式只对支持cmake编译安装的库有效）
set(SOURCE_FILES main.cpp)
include_directories(${OpenCV_INCLUDE_DIRS})add_executable(untitled main.cpp)
target_link_libraries(untitled ${OpenCV_LIBS})

#include <iostream>
#include <opencv2/opencv.hpp>
using namespace std;
using namespace cv;int main() {string path = "/home/dell/下载/a.png";Mat src = imread(path);namedWindow("img");imshow("img",src);waitKey(0);return 0;
}

• 重启一下就ok了

• 结果

onnx

效果风格转换

网络的论文和ONNX下载

• 下载地址
  

blobFromImages(blobFromImage) + imagesFromBlob

        处理图像到blob,[0, 255] ->[0, 1],大小，RGB->BGR和转换forward的blob到图像。

网络传播整体代码

cmake_minimum_required(VERSION 3.15)
project(untitled)
set(CMAKE_CXX_STANDARD 14)MESSAGE(STATUS "Project: untitled")             
find_package(OpenCV REQUIRED)
set(SOURCE_FILES main.cpp)
include_directories(${OpenCV_INCLUDE_DIRS})add_executable(untitled main.cpp)
target_link_libraries(untitled ${OpenCV_LIBS})

#include <opencv2/opencv.hpp>
#include <opencv2/dnn.hpp>
#include <iostream>
#include <fstream>using namespace cv;
using namespace cv::dnn;
using namespace std;// 图像处理  标准化处理
void PreProcess(const Mat& image, Mat& image_blob)
{Mat input;image.copyTo(input);//数据处理 标准化std::vector<Mat> channels, channel_p;split(input, channels);Mat R, G, B;B = channels.at(0);G = channels.at(1);R = channels.at(2);B = (B / 255. - 0.406) / 0.225;G = (G / 255. - 0.456) / 0.224;R = (R / 255. - 0.485) / 0.229;channel_p.push_back(R);channel_p.push_back(G);channel_p.push_back(B);Mat outt;merge(channel_p, outt);image_blob = outt;
}String bin_model = "/home/dell/CLionProjects/untitled/mosaic-9.onnx";
int main(int argc, char** argv) {//数据处理Mat test = Mat::zeros(10,10, CV_64FC1 );Mat image1 = imread("/home/dell/下载/a.png");resize(image1, image1, Size(256, 256), INTER_AREA);image1.convertTo(image1, CV_32FC3);// convertTo()数据类型CV_32FC3PreProcess(image1, image1);vector<Mat> images;images.push_back(image1);images.push_back(image1);int w = 224;int h = 224;// 加载网络cv::dnn::Net net = cv::dnn::readNetFromONNX(bin_model);  // 加载训练好的识别模型  net = cv2.dnn.readNetFromONNX('**.onnx')net.setPreferableBackend(cv::dnn::DNN_BACKEND_OPENCV);net.setPreferableTarget(cv::dnn::DNN_TARGET_CPU);if (net.empty()) {printf("read onnx model data failure...\n");return -1;}Mat inputBlob = blobFromImages(images, 1.0, Size(w, h), Scalar(0, 0, 0), false, true);net.setInput(inputBlob);cv::Mat prob = net.forward();     // 推理出结果  cols,rows 矩阵的行数，列数【注意，在图像中行数对应的是高度，列数对应的是宽度】，当维数大于2时，均为-1;  std::vector<cv::Mat> predTmp;cv::dnn::imagesFromBlob(prob, predTmp);imshow("show Image", images[0]);cv::waitKey(0);imshow("Image mosaic", predTmp[0]);cv::waitKey(0);vector<double> times;double time = net.getPerfProfile(times);float ms = (time * 1000) / getTickFrequency();printf("current inference time : %.2f ms \n", ms);return 0;
}

CG

• clion中debug不生效
  

代码 onnx+ opdncv+vgg16

// https://blog.csdn.net/qq_44747572/article/details/121467657
// /home/dell/下载/opencv-4.6.0/samples/data/dnn/classification_classes_ILSVRC2012.txt#include <opencv2/opencv.hpp>
#include <opencv2/dnn.hpp>
#include <iostream>
#include <fstream>using namespace cv;
using namespace cv::dnn;
using namespace std;// 图像处理  标准化处理
void PreProcess(const Mat& image, Mat& image_blob)
{Mat input;image.copyTo(input);//数据处理 标准化std::vector<Mat> channels, channel_p;split(input, channels);Mat R, G, B;B = channels.at(0);G = channels.at(1);R = channels.at(2);B = (B / 255. - 0.406) / 0.225;G = (G / 255. - 0.456) / 0.224;R = (R / 255. - 0.485) / 0.229;channel_p.push_back(R);channel_p.push_back(G);channel_p.push_back(B);Mat outt;merge(channel_p, outt);image_blob = outt;
}std::vector<String> readClassNames(string labels_txt_file)
{std::vector<String> classNames;std::ifstream fp(labels_txt_file);if (!fp.is_open()){printf("could not open file...\n");exit(-1);}std::string name;while (!fp.eof()){std::getline(fp, name);if (name.length())classNames.push_back(name);}fp.close();return classNames;
}String bin_model = "/home/dell/下载/vgg16.onnx";
String labels_txt_file = "/home/dell/下载/opencv-4.6.0/samples/data/dnn/classification_classes_ILSVRC2012.txt";
vector<String> readClassNames();                  // string对象作为vector对象
int main(int argc, char** argv) {Mat image1 = imread("/home/dell/下载/a.png");//用于显示vector<Mat>Showimages;Showimages.push_back(image1);//处理image1resize(image1, image1, Size(256, 256), INTER_AREA);image1.convertTo(image1, CV_32FC3);PreProcess(image1, image1);         //标准化处理//将image1和image2合并到imagesvector<Mat> images;images.push_back(image1);images.push_back(image1);vector<String> labels = readClassNames(labels_txt_file);int w = 224;int h = 224;// 加载网络cv::dnn::Net net = cv::dnn::readNetFromONNX(bin_model);  // 加载训练好的识别模型if (net.empty()) {printf("read onnx model data failure...\n");return -1;}Mat inputBlob = blobFromImages(images, 1.0, Size(w, h), Scalar(0, 0, 0), false, true);// 执行图像分类net.setInput(inputBlob);cv::Mat prob = net.forward();     // 推理出结果cout << prob.cols<< endl;vector<double> times;double time = net.getPerfProfile(times);float ms = (time * 1000) / getTickFrequency();printf("current inference time : %.2f ms \n", ms);// 得到最可能分类输出for (int n = 0; n < prob.rows; n++) {Point classNumber;double classProb;Mat probMat = prob(Rect(0, n, 1000, 1)).clone();Mat result = probMat.reshape(1, 1);minMaxLoc(result, NULL, &classProb, NULL, &classNumber);int classidx = classNumber.x;printf("\n current image classification : %s, possible : %.2f\n", labels.at(classidx).c_str(), classProb);// 显示文本putText(Showimages[n], labels.at(classidx), Point(10, 20), FONT_HERSHEY_SIMPLEX, 0.6, Scalar(0, 0, 255), 1, 1);imshow("Image Classification", Showimages[n]);waitKey(0);}return 0;
}