NanjingForestryUniversity/cotton_color
3
1
Fork 0
mirror of https://github.com/Karllzy/cotton_color.git synced 2025-11-09 11:13:54 +00:00
Code Issues Actions Packages Projects Releases Wiki Activity

添加了串行执行的示例

Browse Source
This commit is contained in:
zjc-zjc-123 2024-11-27 13:39:29 +08:00
parent 50058f4ad1
commit d1e0320d29
9 changed files with 161 additions and 141 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

6
opencv_onnx.cpp
View File

@ -124,7 +124,7 @@ int main() {
timer1.printElapsedTime("Time to preprocess image");
for (int j = 0; j < 1; j++) {
for (int j = 0; j < 10; j++) {
// Run inference
cv::Mat output = net.forward();
@ -187,8 +187,8 @@ int main() {
cv::imshow("Detection Mask", detectionMask);
// Save the result as an image
std::string savepath = "C:\\Users\\zjc\\Desktop\\suspect_mask.png";
cv::imwrite(savepath, detectionMask);
// std::string savepath = "C:\\Users\\zjc\\Desktop\\suspect_mask.png";
// cv::imwrite(savepath, detectionMask);
timer1.printElapsedTime("Time to run inference");
}

3
src/CMakeLists.txt
View File

@ -11,7 +11,6 @@ target_include_directories(Matrox PUBLIC ${CMAKE_CURRENT_SOURCE_DIR})
target_link_libraries(Matrox PUBLIC Qt6::Widgets ${OpenCV_LIBS} ${MIL_LIBS})
#定义CVDL模块的库
add_library(CVDL
CVDL/OnnxRunner.cpp
@ -19,4 +18,4 @@ add_library(CVDL
# 头文件路径
target_include_directories(CVDL PUBLIC ${CMAKE_CURRENT_SOURCE_DIR})
# 链接依赖库
target_link_libraries(CVDL PUBLIC Qt6::Widgets ${OpenCV_LIBS} ${MIL_LIBS})
target_link_libraries(CVDL PUBLIC Qt6::Widgets Matrox ${OpenCV_LIBS} ${MIL_LIBS})

40
src/CVDL/OnnxRunner.cpp
View File

@ -15,7 +15,7 @@ void Timer::printElapsedTime(const std::string& message) {
}
// Resize and pad input image
cv::Mat resizeAndPad(const cv::Mat& image, int targetWidth, int targetHeight, int& padTop, int& padLeft, float& scale, const cv::Scalar& padColor) {
cv::Mat ONNXRunner::resizeAndPad(const cv::Mat& image, int targetWidth, int targetHeight, int& padTop, int& padLeft, float& scale, const cv::Scalar& padColor) {
int originalWidth = image.cols;
int originalHeight = image.rows;
@ -38,7 +38,7 @@ cv::Mat resizeAndPad(const cv::Mat& image, int targetWidth, int targetHeight, in
}
// Create detection mask
cv::Mat createDetectionMask(const cv::Mat& originalImage, const std::vector<Detection>& detections, float scale, int padTop, int padLeft) {
cv::Mat ONNXRunner::createDetectionMask(const cv::Mat& originalImage, const std::vector<Detection>& detections, float scale, int padTop, int padLeft) {
cv::Mat mask = cv::Mat::zeros(originalImage.size(), CV_8UC1); // Single channel mask
for (const auto& detection : detections) {
@ -59,7 +59,7 @@ cv::Mat createDetectionMask(const cv::Mat& originalImage, const std::vector<Dete
}
// Load the ONNX model
cv::dnn::Net loadModel(const std::string& modelPath) {
cv::dnn::Net ONNXRunner::loadModel(const std::string& modelPath) {
cv::dnn::Net net = cv::dnn::readNetFromONNX(modelPath);
net.setPreferableBackend(cv::dnn::DNN_BACKEND_CUDA); // Use CUDA backend
net.setPreferableTarget(cv::dnn::DNN_TARGET_CUDA); // Run on GPU
@ -67,7 +67,7 @@ cv::dnn::Net loadModel(const std::string& modelPath) {
}
// Preprocess image for model input
cv::Mat preprocessImage(const cv::Mat& image, cv::dnn::Net& net, int& padTop, int& padLeft, float& scale) {
cv::Mat ONNXRunner::preprocessImage(const cv::Mat& image, cv::dnn::Net& net, int& padTop, int& padLeft, float& scale) const {
cv::Scalar padColor(128, 128, 128); // Gray padding
cv::Mat inputImage = resizeAndPad(image, INPUT_WIDTH, INPUT_HEIGHT, padTop, padLeft, scale, padColor);
cv::Mat blob = cv::dnn::blobFromImage(inputImage, 1 / 255.0, cv::Size(INPUT_WIDTH, INPUT_HEIGHT), cv::Scalar(0, 0, 0), true, false);
@ -76,7 +76,7 @@ cv::Mat preprocessImage(const cv::Mat& image, cv::dnn::Net& net, int& padTop, in
}
// Perform inference on the input image
std::vector<Detection> performInference(cv::dnn::Net& net, const cv::Mat& inputImage) {
std::vector<Detection> ONNXRunner::performInference(cv::dnn::Net& net, const cv::Mat& inputImage) const {
std::vector<Detection> detections;
cv::Mat output = net.forward();
float* data = (float*)output.data;
@ -112,7 +112,7 @@ std::vector<Detection> performInference(cv::dnn::Net& net, const cv::Mat& inputI
}
// Apply Non-Maximum Suppression
std::vector<Detection> applyNMS(std::vector<Detection>& detections) {
std::vector<Detection> ONNXRunner::applyNMS(std::vector<Detection>& detections) const {
std::vector<int> indices;
std::vector<cv::Rect> boxes;
std::vector<float> scores;
@ -130,3 +130,31 @@ std::vector<Detection> applyNMS(std::vector<Detection>& detections) {
return finalDetections;
}
void ONNXRunner::load(const std::string& modelPath) {
this->net = ONNXRunner::loadModel(modelPath);
// 创建一个空的输入矩阵作为预热数据（假定模型输入是 RGB 图像）
cv::Mat dummyInput = cv::Mat::zeros(INPUT_HEIGHT, INPUT_WIDTH, CV_8UC3); // 640x640 的全零矩阵
// 调用 predict 方法进行预热
this->predict(dummyInput);
}; // Load the model
std::vector<Detection> ONNXRunner::predict(const cv::Mat &image) {
cv::dnn::Net net = this->net;
cv::Mat inputImage = preprocessImage(image, net, this->pad_top, this->pad_left, this->scale);
// Perform inference
std::vector<Detection> detections = performInference(net, inputImage);
// Apply Non-Maximum Suppression
std::vector<Detection> finalDetections = applyNMS(detections);
return finalDetections;
}
cv::Mat ONNXRunner::postProcess(const std::vector<Detection>& detections, const cv::Mat& image) const {
// Create and show the detection mask
cv::Mat detectionMask = ONNXRunner::createDetectionMask(image, detections, this->scale, this->pad_top, this->pad_left);
return detectionMask;
}

32
src/CVDL/OnnxRunner.h
View File

@ -27,13 +27,31 @@ public:
private:
std::chrono::high_resolution_clock::time_point start_time;
};
class ONNXRunner {
public:
ONNXRunner() = default; // Default constructor
void load(const std::string& modelPath); // Load the model
std::vector<Detection> predict(const cv::Mat& image); // Predict and generate results
cv::Mat postProcess(const std::vector<Detection> &detections, const cv::Mat &image) const;
private:
cv::dnn::Net net;
int pad_top, pad_left;
float scale;
// Function prototypes
static cv::Mat resizeAndPad(const cv::Mat& image, int targetWidth, int targetHeight, int& padTop, int& padLeft, float& scale, const cv::Scalar& padColor);
static cv::Mat createDetectionMask(const cv::Mat& originalImage, const std::vector<Detection>& detections, float scale, int padTop, int padLeft);
static cv::dnn::Net loadModel(const std::string& modelPath);
cv::Mat preprocessImage(const cv::Mat& image, cv::dnn::Net& net, int& padTop, int& padLeft, float& scale) const;
std::vector<Detection> performInference(cv::dnn::Net& net, const cv::Mat& inputImage) const;
std::vector<Detection> applyNMS(std::vector<Detection>& detections) const;
// Constants
const float CONFIDENCE_THRESHOLD = 0.2f;
const float NMS_THRESHOLD = 0.2f;
const int INPUT_WIDTH = 640;
const int INPUT_HEIGHT = 640;
};
// Function prototypes
cv::Mat resizeAndPad(const cv::Mat& image, int targetWidth, int targetHeight, int& padTop, int& padLeft, float& scale, const cv::Scalar& padColor);
cv::Mat createDetectionMask(const cv::Mat& originalImage, const std::vector<Detection>& detections, float scale, int padTop, int padLeft);
cv::dnn::Net loadModel(const std::string& modelPath);
cv::Mat preprocessImage(const cv::Mat& image, cv::dnn::Net& net, int& padTop, int& padLeft, float& scale);
std::vector<Detection> performInference(cv::dnn::Net& net, const cv::Mat& inputImage);
std::vector<Detection> applyNMS(std::vector<Detection>& detections);
#endif // ONNXRUNNER_H

53
src/Matrox/template_matching.cpp
View File

@ -119,8 +119,9 @@ void TemplateMatcher::findModels(const MIL_ID& inputImage,MIL_ID& outputImage)
return;
}
MIL_ID input_image_uint8 = convert_to_uint8(inputImage);
MdispSelect(MilDisplay, input_image_uint8);
if(this->param["isdisplay"] == 1) {
MdispSelect(MilDisplay, input_image_uint8);
}
// Clear previous annotations
MgraClear(M_DEFAULT, GraphicList);
@ -155,13 +156,16 @@ void TemplateMatcher::findModels(const MIL_ID& inputImage,MIL_ID& outputImage)
MbufClear(outputImage, 0);
// Display results
std::cout << "Found " << NumResults << " model(s) in " << Time * 1000.0 << " ms:\n";
std::cout << "Result Model X Position Y Position Angle Scale Score\n";
if(this->param["debug_mode"] == 1) {
std::cout << "Found " << NumResults << " model(s) in " << Time * 1000.0 << " ms:\n";
std::cout << "Result Model X Position Y Position Angle Scale Score\n";
}
for (MIL_INT i = 0; i < NumResults; ++i) {
std::cout << i << " " << Models[i] << " " << XPosition[i] << " "
<< YPosition[i] << " " << Angle[i] << " " << Scale[i]
<< " " << Score[i] << "%\n";
if (this->param["debug_mode"] == 1) {
std::cout << i << " " << Models[i] << " " << XPosition[i] << " "
<< YPosition[i] << " " << Angle[i] << " " << Scale[i]
<< " " << Score[i] << "%\n";
}
// Draw results onto the binary image
MgraColor(M_DEFAULT, 255); // White color for binary image
MmodDraw(M_DEFAULT, MilResult, outputImage, M_DRAW_EDGES + M_DRAW_POSITION, i, M_DEFAULT);
@ -172,15 +176,11 @@ void TemplateMatcher::findModels(const MIL_ID& inputImage,MIL_ID& outputImage)
M_DRAW_EDGES + M_DRAW_POSITION, i, M_DEFAULT);
}
// Display or save the binary image
MbufSave(SAVE_PATH2, outputImage);
} else {
std::cout << "No models found.\n";
}
MosPrintf(MIL_TEXT("Press <Enter> to EXIT.\n\n"));
MosGetch();
// MosPrintf(MIL_TEXT("Press <Enter> to EXIT.\n\n"));
// MosGetch();
MbufFree(input_image_uint8);
}
@ -237,15 +237,14 @@ void TemplateMatcher::loadConfig(const std::string& filename,
file.close();
}
void TemplateMatcher::LoadTemplate(std::map<std::string, int>& params)
void TemplateMatcher::LoadConfig(const string& config_path)
{
std::vector<std::string> template_paths;
std::vector<MIL_INT> offsetX, offsetY, sizeX, sizeY;
std::vector<MIL_DOUBLE> drawColor;
// 调用 loadConfig 并加载配置
loadConfig("C:\\Users\\zjc\\Desktop\\config\\template_config.txt",
template_paths, offsetX, offsetY, sizeX, sizeY, drawColor);
loadConfig(config_path, template_paths, offsetX, offsetY, sizeX, sizeY, drawColor);
// 调用 matcher 的 loadTemplates 方法
this->loadTemplates(template_paths, offsetX, offsetY, sizeX, sizeY, drawColor);
@ -256,12 +255,24 @@ void TemplateMatcher::FindTemplates( const MIL_ID& inputImage, MIL_ID& outputIma
{
// Perform template matching
this -> findModels(inputImage,outputImage);
// Notify user that matching is complete
cout << "Template matching completed.\n";
}
void TemplateMatcher::predict(const MIL_ID &inputImage, MIL_ID &outputImage, const std::map<std::string, int> &params) {
MIL_ID detection_result, detection_resize, output_image_unresize;
pre_process(inputImage, detection_result, params);
MbufAlloc2d(MilSystem, MbufInquire(detection_result, M_SIZE_X, M_NULL)/2,
MbufInquire(detection_result, M_SIZE_Y, M_NULL)/2, 1 + M_UNSIGNED,
M_IMAGE + M_PROC, &detection_resize);
MimResize(detection_result,detection_resize,0.5,0.5,M_DEFAULT);
this->FindTemplates(detection_resize,output_image_unresize, params);
MbufAlloc2d(MilSystem, MbufInquire(detection_result, M_SIZE_X, M_NULL),
MbufInquire(detection_result, M_SIZE_Y, M_NULL), 1 + M_UNSIGNED,
M_IMAGE + M_PROC, &outputImage);
MimResize(output_image_unresize,outputImage,2,2,M_DEFAULT);
MbufFree(detection_result);
MbufFree(detection_resize);
MbufFree(output_image_unresize);
}
// TODO: Opencv ONNX runner, 已

9
src/Matrox/template_matching.h
View File

@ -21,7 +21,6 @@ private:
MIL_ID MilSearchContext;
MIL_ID MilResult;
MIL_ID GraphicList;
#define SAVE_PATH2 MIL_TEXT("C:\\Users\\zjc\\Desktop\\detection.png")
std::vector<std::string> ModelImgPaths;
std::vector<MIL_INT> ModelsOffsetX;
@ -38,7 +37,7 @@ public:
// Constructor
TemplateMatcher(MIL_ID system, MIL_ID display, std::map<std::string, int>& param);
void LoadConfig(const std::string& config_path);
// Load template models
void loadTemplates(const std::vector<std::string>& template_paths,
@ -49,11 +48,13 @@ public:
// Search for models in the input image
void findModels(const MIL_ID& inputImage,MIL_ID& outputImage);
void LoadTemplate(std::map<std::string, int> &params);
void FindTemplates(const MIL_ID &inputImage, MIL_ID &outputImage,const std::map<std::string, int> &params);
void loadConfig(const std::string& filename,
void predict(const MIL_ID& inputImage, MIL_ID& outputImage, const std::map<std::string, int> &params);
static void loadConfig(const std::string& filename,
std::vector<std::string>& template_paths,
std::vector<MIL_INT>& offsetX,
std::vector<MIL_INT>& offsetY,

36
tests/test_onnx.cpp
View File

@ -7,7 +7,8 @@ int main() {
Timer timer1;
// Load the model
cv::dnn::Net net = loadModel(modelPath);
ONNXRunner runner;
runner.load(modelPath);
timer1.printElapsedTime("Time to load the model");
// Read the input image
@ -16,30 +17,19 @@ int main() {
std::cerr << "Could not read the image: " << imagePath << std::endl;
return -1;
}
// Preprocess image
int padTop, padLeft;
float scale;
cv::Mat inputImage = preprocessImage(image, net, padTop, padLeft, scale);
timer1.printElapsedTime("Time to preprocess image");
// Perform inference
std::vector<Detection> detections = performInference(net, inputImage);
// Apply Non-Maximum Suppression
std::vector<Detection> finalDetections = applyNMS(detections);
std::cout << "Number of detections after NMS: " << finalDetections.size() << std::endl;
// Create and show the detection mask
cv::Mat detectionMask = createDetectionMask(image, finalDetections, scale, padTop, padLeft);
// cv::imshow("Detection Mask", detectionMask);
cv::Mat mask;
timer1.printElapsedTime("Time to load image.");
for(int i = 0; i < 10; i++) {
std:: cout << "Run time: " << i << std::endl;
std::vector<Detection> result = runner.predict(image);
timer1.printElapsedTime("Time to predict result");
mask = runner.postProcess(result, image);
timer1.printElapsedTime("Time to predict result");
}
// Save the result
std::string savepath = "C:\\Users\\zjc\\Desktop\\suspect_mask.png";
cv::imwrite(savepath, detectionMask);
timer1.printElapsedTime("Time to run inference");
cv::imwrite(savepath, mask);
timer1.printElapsedTime("Time to save image");
cv::waitKey(0);
return 0;

4
tests/test_template_matching.cpp
View File

@ -37,7 +37,7 @@ int main() {
MIL_ID output_Image= M_NULL;
TemplateMatcher matcher(MilSystem, MilDisplay, params);
matcher.LoadConfig("C:\\Users\\zjc\\Desktop\\config\\template_config.txt");
// Measure execution time
measure_execution_time([&]()
{
@ -46,8 +46,6 @@ int main() {
MbufInquire(detection_result, M_SIZE_Y, M_NULL)/2, 1 + M_UNSIGNED,
M_IMAGE + M_PROC, &detection_resize);
MimResize(detection_result,detection_resize,0.5,0.5,M_DEFAULT);
matcher.LoadTemplate(params);
matcher.FindTemplates(detection_resize,output_Image,params);
});

119
tests/test_total.cpp
View File

@ -2,6 +2,8 @@
#include <iostream>
#include <map>
#include <string>
#include <CVDL/OnnxRunner.h>
#include "Mil.h"
#include "Matrox/utils.h"
#include "Matrox/color_range.h"
@ -10,101 +12,74 @@
// 宏定义
#define SAVE_PATH3 MIL_TEXT("C:\\Users\\zjc\\Desktop\\suspect.png")
#define SAVE_PATH4 MIL_TEXT("C:\\Users\\zjc\\Desktop\\suspect2.png")
#define IMAGE_PATH MIL_TEXT("C:\\Users\\zjc\\Desktop\\cotton_image_new\\357.bmp")
#define IMAGE_PATH MIL_TEXT(".\\test_imgs\\357.bmp")
std::string imagePath = "C:\\Users\\zjc\\Desktop\\dimo.bmp";
MIL_ID MilApplication, MilSystem, MilDisplay;
std::map<std::string, int> params;
void run_high_sat_detect(MIL_ID MilImage, std::map<std::string, int>& params) {
MIL_ID detection_result = M_NULL;
read_params_from_file("C:\\Users\\zjc\\Desktop\\config\\color_range_config.txt", params);
measure_execution_time([&]() {
high_sat_detect(MilImage, detection_result, params);
});
MbufSave(SAVE_PATH3, detection_result);
MbufFree(detection_result);
}
void run_pre_process(MIL_ID MilImage, std::map<std::string, int>& params) {
MIL_ID detection_result = M_NULL;
MIL_ID detection_resize = M_NULL;
MIL_ID output_Image = M_NULL;
read_params_from_file("C:\\Users\\zjc\\Desktop\\config\\template_color_config.txt", params);
TemplateMatcher matcher(MilSystem, MilDisplay, params);
measure_execution_time([&]() {
pre_process(MilImage, detection_result, params);
MbufAlloc2d(MilSystem, MbufInquire(detection_result, M_SIZE_X, M_NULL) / 2,
MbufInquire(detection_result, M_SIZE_Y, M_NULL) / 2, 1 + M_UNSIGNED,
M_IMAGE + M_PROC, &detection_resize);
MimResize(detection_result, detection_resize, 0.5, 0.5, M_DEFAULT);
matcher.LoadTemplate(params);
matcher.FindTemplates(detection_resize, output_Image, params);
});
MbufSave(SAVE_PATH4, detection_result);
MbufFree(detection_resize);
MbufFree(output_Image);
MbufFree(detection_result);
}
using namespace std;
int main() {
// 初始化 MIL 应用
MappAllocDefault(M_DEFAULT, &MilApplication, &MilSystem, &MilDisplay, M_NULL, M_NULL) ;
// 读取图片
MIL_ID MilImage = M_NULL;
MIL_ID MilImage=M_NULL, MilHighSatResult = M_NULL, MilTemplateMatchingResult = M_NULL;
MbufRestore(IMAGE_PATH, MilSystem, &MilImage);
if (MilImage == M_NULL) {
std::cerr << "Error: Failed to restore image from " << IMAGE_PATH << std::endl;
MappFreeDefault(MilApplication, MilSystem, MilDisplay, M_NULL, M_NULL);
return -1;
}
Timer timer1, timer2;
std::map<std::string, int> params;
read_params_from_file("..\\config\\color_range_config.txt", params);
read_params_from_file("..\\config\\template_color_config.txt", params);
TemplateMatcher matcher(MilSystem, MilDisplay, params);
matcher.LoadConfig("..\\config\\template_config.txt");
ONNXRunner runner;
runner.load("C:\\Users\\zjc\\Desktop\\dimo_11.14.onnx");
timer1.printElapsedTime("Load config and templates and models");
cv::Mat deep_result, high_sat_result, template_result, total_result;
cout << "Sequence running start:" << endl;
for(int i = 0; i < 10; i++) {
cout << i << endl;
timer1.restart();
timer2.restart();
// 艳丽色彩检测
high_sat_detect(MilImage, MilHighSatResult, params);
timer1.printElapsedTime("High Sat Predict finished");
// 模板匹配检测
matcher.predict(MilImage, MilTemplateMatchingResult, params);
timer1.printElapsedTime("Template Matching Predict finished");
// 深度学习检测
cv::Mat cv_input = mil2mat(MilImage);
std::vector<Detection> result = runner.predict(cv_input);
deep_result = runner.postProcess(result, cv_input);
std::string savepath = "C:\\Users\\zjc\\Desktop\\suspect_mask2.png";
cv::imwrite(savepath, deep_result);
timer1.printElapsedTime("Deep Learning Predict finished");
high_sat_result = mil2mat(MilHighSatResult);
template_result = mil2mat(MilTemplateMatchingResult);
// 加载参数
// 创建两个子进程
PROCESS_INFORMATION processInfo1, processInfo2;
STARTUPINFO startupInfo1, startupInfo2;
if (!deep_result.empty() && !high_sat_result.empty() && !template_result.empty()) {
cv::bitwise_and(deep_result, high_sat_result, total_result);
cv::bitwise_and(total_result, template_result, total_result);
} else {
cerr << "Error: One or more detection results are empty!" << endl;
}
timer2.printElapsedTime("Prediction finished Total");
ZeroMemory(&startupInfo1, sizeof(startupInfo1));
ZeroMemory(&startupInfo2, sizeof(startupInfo2));
ZeroMemory(&processInfo1, sizeof(processInfo1));
ZeroMemory(&processInfo2, sizeof(processInfo2));
cv::imshow("Result", total_result);
cv::waitKey(0);
startupInfo1.cb = sizeof(startupInfo1);
startupInfo2.cb = sizeof(startupInfo2);
// 子进程1
if (!CreateProcess(NULL, LPWSTR((LPSTR) "ChildProcess1.exe"), NULL, NULL, FALSE, 0, NULL, NULL, &startupInfo1, &processInfo1)) {
std::cerr << "Error: Failed to create process 1. Error code: " << GetLastError() << std::endl;
MbufFree(MilImage);
MappFreeDefault(MilApplication, MilSystem, MilDisplay, M_NULL, M_NULL);
return -1;
}
// 子进程2
if (!CreateProcess(NULL, LPWSTR((LPSTR) "ChildProcess2.exe"), NULL, NULL, FALSE, 0, NULL, NULL, &startupInfo2, &processInfo2)) {
std::cerr << "Error: Failed to create process 2. Error code: " << GetLastError() << std::endl;
MbufFree(MilImage);
MappFreeDefault(MilApplication, MilSystem, MilDisplay, M_NULL, M_NULL);
return -1;
}
// 等待两个子进程完成
WaitForSingleObject(processInfo1.hProcess, INFINITE);
WaitForSingleObject(processInfo2.hProcess, INFINITE);
CloseHandle(processInfo1.hProcess);
CloseHandle(processInfo1.hThread);
CloseHandle(processInfo2.hProcess);
CloseHandle(processInfo2.hThread);
// 释放资源
MbufFree(MilImage);
MbufFree(MilHighSatResult);
MbufFree(MilTemplateMatchingResult);
MappFreeDefault(MilApplication, MilSystem, MilDisplay, M_NULL, M_NULL);
std::cout << "所有模块检测已完成！按 <Enter> 退出。" << std::endl;