本文opencv版本3.3（3.3以上版本支持dnn）
經過測試opencv調用深度學習比caffe提供的C++接口效率還要高一些
分類及目標檢測代碼如下：
#include <string>
#include <unistd.h>
#include <stdlib.h>
#include <sys/types.h>
#include <sys/stat.h>
#include <fcntl.h>
#include <string.h>
#include <iostream>
#include "opencv/cv.h"
#include "opencv/highgui.h"
#include "opencv/cvwimage.h"
#include <opencv2/opencv.hpp>
#include <opencv2/imgproc/imgproc.hpp>
#include <opencv2/dnn.hpp>
#include <opencv2/dnn/shape_utils.hpp>
#include <opencv2/core/utils/trace.hpp>
#include <opencv2/core/core.hpp>
#include <opencv2/core/utility.hpp>
using namespace cv;
using namespace cv::dnn;
const size_t width = 300;
const size_t height = 300;
//尋找出概率最高的一類
static void getMaxClass(const Mat &probBlob, int *classId, double *classProb)
{
Mat probMat = probBlob.reshape(1, 1);
Point classNumber;
minMaxLoc(probMat, NULL, classProb, NULL, &classNumber);
*classId = classNumber.x;
}
//從標籤文件讀取分類 空格爲標誌
static std::vector<String> readClassNames(const char *filename = "labels.txt")
{
std::vector<String> classNames;
std::ifstream fp(filename);
if (!fp.is_open())
{
std::cerr << "File with classes labels not found: " << filename << std::endl;
exit(-1);
}
std::string name;
while (!fp.eof())
{
std::getline(fp, name);
if (name.length())
classNames.push_back(name.substr(name.find(' ') + 1));
}
fp.close();
return classNames;
}
static std::string classification(String modelTxt, String modelBin,String imageFile,const char *lablesFile,cv::Rect tRect)
{
//初始化
CV_TRACE_FUNCTION();
//讀取模型參數和模型結構文件
//modelTxt = "caffenet.prototxt";
//modelBin = "caffenet.caffemodel";
//讀取圖片
//imageFile = "test0.jpg";
//合成網絡
Net net = dnn::readNetFromCaffe(modelTxt, modelBin);
//判斷網絡是否生成成功
if (net.empty())
{
LOGE("Can't load network by using the following files: ");
return NULL;
}
LOGE("net read successfully");
//讀取圖片
Mat img1 = imread(imageFile);
Mat img = img1 (cv::Rect(tRect.x, tRect.y, tRect.width, tRect.height));
//imshow("image", img);
if (img.empty())
{
std::cerr << "Can't read image from the file: " << imageFile << std::endl;
return NULL;
}
LOGE("image read sucessfully");
/* Mat inputBlob = blobFromImage(img, 1, Size(224, 224),
Scalar(104, 117, 123)); */
//構造blob，爲傳入網絡做準備，圖片不能直接進入網絡
Mat inputBlob = blobFromImage(img, 1, Size(224, 224));
Mat prob;
cv::TickMeter t;
for (int i = 0; i < 10; i++)//此處可以只循環一次
{
CV_TRACE_REGION("forward");
//將構建的blob傳入網絡data層
net.setInput(inputBlob,"data");
//計時
t.start();
//前向預測
prob = net.forward("softmax");
//停止計時
t.stop();
}
int classId;
double classProb;
//找出最高的概率ID存儲在classId，對應的標籤在classProb中
getMaxClass(prob, &classId, &classProb);
//打印出結果
std::vector<String> classNames = readClassNames(lablesFile);
LOGE("Best class: #%d",classId);
LOGE("class name %s",classNames.at(classId).c_str());
LOGE("Probability: %d",(int)classProb * 100);
//打印出花費時間
LOGE("Time: %f ms",(double)t.getTimeMilli() / t.getCounter());
return classNames.at(classId).c_str();
}
static Mat getMean(const size_t& imageHeight, const size_t& imageWidth)
{
Mat mean;
const int meanValues[3] = {104, 117, 123};
std::vector<Mat> meanChannels;
for(int i = 0; i < 3; i++)
{
Mat channel((int)imageHeight, (int)imageWidth, CV_32F, Scalar(meanValues[i]));
meanChannels.push_back(channel);
}
cv::merge(meanChannels, mean);
return mean;
}
static Mat preprocess(const Mat& frame)
{
Mat preprocessed;
frame.convertTo(preprocessed, CV_32F);
resize(preprocessed, preprocessed, Size(width, height)); //SSD accepts 300x300 RGB-images
Mat mean = getMean(width, height);
cv::subtract(preprocessed, mean, preprocessed);
return preprocessed;
}
static cv::Rect ssdCheck(String modelTxt, String modelBinary,String imageFile)
{
cv::Rect dstObject(0,0,0,0);
//讀取圖片
Mat srcImage = imread(imageFile);
Ptr<dnn::Importer> importer;
try {
importer = dnn::createCaffeImporter(modelTxt,modelBinary);
}
catch (const cv::Exception &err) {
LOGE(" import caffe model error");
}
if (!importer)
{
LOGE("Can't load network by using the following files:");
LOGE("/mnt/sdcard/ssd.caffemodel /mnt/sdcard/ssd.prototxt");
return dstObject;
}
//! [Initialize network]
dnn::Net net;
importer->populateNet(net);
importer.release(); //We don't need importer anymore
//! [Initialize network]
cv::Mat frame = srcImage;
if (frame.channels() == 4)
cvtColor(frame, frame, COLOR_BGRA2BGR);
//! [Prepare blob]
Mat preprocessedFrame = preprocess(frame);
Mat inputBlob = blobFromImage(preprocessedFrame); //Convert Mat to batch of images
//! [Prepare blob]
//! [Set input blob]
net.setInput(inputBlob, "data"); //set the network input
//! [Set input blob]
TickMeter tm;
tm.start();
//! [Make forward pass]
Mat detection = net.forward("detection_out"); //compute output
//! [Make forward pass]
tm.stop();
LOGE("process time %f",tm.getTimeMilli());
Mat detectionMat(detection.size[2], detection.size[3], CV_32F, detection.ptr<float>());
float confidenceThreshold = 0.5;
for(int i = 0; i < detectionMat.rows; i++)
{
float confidence = detectionMat.at<float>(i, 2);
if(confidence > confidenceThreshold)
{
size_t objectClass = (size_t)(detectionMat.at<float>(i, 1));
float xLeftBottom = detectionMat.at<float>(i, 3) * frame.cols;
float yLeftBottom = detectionMat.at<float>(i, 4) * frame.rows;
float xRightTop = detectionMat.at<float>(i, 5) * frame.cols;
float yRightTop = detectionMat.at<float>(i, 6) * frame.rows;
// std::cout << " " << xLeftBottom
// << " " << yLeftBottom
// << " " << xRightTop
// << " " << yRightTop << std::endl;
dstObject.x=(int)xLeftBottom;
dstObject.y=(int)yLeftBottom;
dstObject.width=(int)(xRightTop - xLeftBottom);
dstObject.height=(int)(yRightTop - yLeftBottom);
}
}
return dstObject;
}
JNIEXPORT jlong JNICALL Java_com_deepTest
(JNIEnv *env, jobject obj, jlong srcImg, jlong dstImg) {
Mat *srcImage = (Mat *) srcImg;
Mat *dstImage = (Mat *) dstImg;
String modelConfiguration = "/mnt/sdcard/ssd.prototxt";
String modelBinary = "/mnt/sdcard/ssd.caffemodel";
Ptr<dnn::Importer> importer;
try {
importer = dnn::createCaffeImporter(modelConfiguration,modelBinary);
}
catch (const cv::Exception &err) {
LOGE("import caffe model error");
}
if (!importer)
{
LOGE("Can't load network by using the following files:");
LOGE("/mnt/sdcard/ssd.caffemodel /mnt/sdcard/ssd.prototxt");
return (jlong)srcImage;
}
//! [Initialize network]
dnn::Net net;
importer->populateNet(net);
importer.release(); //We don't need importer anymore
//! [Initialize network]
cv::Mat frame = *srcImage;
if (frame.channels() == 4)
cvtColor(frame, frame, COLOR_BGRA2BGR);
//! [Prepare blob]
Mat preprocessedFrame = preprocess(frame);
Mat inputBlob = blobFromImage(preprocessedFrame); //Convert Mat to batch of images
//! [Prepare blob]
//! [Set input blob]
net.setInput(inputBlob, "data"); //set the network input
//! [Set input blob]
TickMeter tm;
tm.start();
//! [Make forward pass]
Mat detection = net.forward("detection_out"); //compute output
//! [Make forward pass]
tm.stop();
LOGE("process time %f",tm.getTimeMilli());
Mat detectionMat(detection.size[2], detection.size[3], CV_32F, detection.ptr<float>());
float confidenceThreshold = 0.5;
for(int i = 0; i < detectionMat.rows; i++)
{
float confidence = detectionMat.at<float>(i, 2);
if(confidence > confidenceThreshold)
{
size_t objectClass = (size_t)(detectionMat.at<float>(i, 1));
float xLeftBottom = detectionMat.at<float>(i, 3) * frame.cols;
float yLeftBottom = detectionMat.at<float>(i, 4) * frame.rows;
float xRightTop = detectionMat.at<float>(i, 5) * frame.cols;
float yRightTop = detectionMat.at<float>(i, 6) * frame.rows;
// std::cout << " " << xLeftBottom
// << " " << yLeftBottom
// << " " << xRightTop
// << " " << yRightTop << std::endl;
Rect object((int)xLeftBottom, (int)yLeftBottom,
(int)(xRightTop - xLeftBottom),
(int)(yRightTop - yLeftBottom));
rectangle(frame, object, Scalar(0, 255, 0));
}
}
frame.copyTo(*dstImage);
return (jlong)dstImage;
}