【opencv】神经网络识别美女

最近比较闲，想做一个判断是否是美女的算法
从网上搜集了一些图片，首先要提取这些图片中的人脸并保存作为训练集，可以参考文章：
http://blog.csdn.net/qq_15947787/article/details/51393030

下面是完整的代码

//opencv2.4.9 + vs2012 + win7 x64#include <opencv2/opencv.hpp>#include <iostream>#include <stdio.h>#include <windows.h>using namespace std;using namespace cv;/** Function Headers */void detectAndCut( Mat img ,string dir ,string filename );void AllImagePro( string src, string dst, const int number );char* WcharToChar(const wchar_t* wp);wchar_t* CharToWchar(const char* c);wchar_t* StringToWchar(const string& s);string getstring ( const int n );CascadeClassifier face_cascade;String face_cascade_name = "haarcascade_frontalface_alt.xml";//主函数int main(){    const int sample_mun_perclass = 12;//训练每类图片数量    const int class_mun = 2;//训练类数 一类是美女，一类是丑女 ^-^    const int image_cols = 30;    const int image_rows = 30;    string  fileReadName,fileReadPath;    float trainingData[class_mun*sample_mun_perclass][image_rows*image_cols] = {{0}};//每一行一个训练样本    float labels[class_mun*sample_mun_perclass][class_mun]={{0}};//训练样本标签    if( !face_cascade.load( face_cascade_name ) ){ printf("--(!)Error loading\n"); return -1; };    AllImagePro( "0", "0cut" ,sample_mun_perclass);    AllImagePro( "1", "1cut" ,sample_mun_perclass);    cout<<"cut……OK！"<<endl;     for(int i=0;i<class_mun;++i)//不同类    {        //读取每个类文件夹下所有图像        int j = 0;//每一类读取图像个数计数        fileReadPath = getstring(i) + "cut/" + "*.*";        cout<<"文件夹"<<i<<endl;        HANDLE hFile;        LPCTSTR lpFileName = StringToWchar(fileReadPath);//指定搜索目录和文件类型，如搜索d盘的音频文件可以是"D:\\*.mp3"        WIN32_FIND_DATA pNextInfo;  //搜索得到的文件信息将储存在pNextInfo中;        hFile = FindFirstFile(lpFileName,&pNextInfo);//请注意是 &pNextInfo , 不是 pNextInfo;        if(hFile == INVALID_HANDLE_VALUE)        {            exit(-1);//搜索失败        }        //do-while循环读取        do        {            if(pNextInfo.cFileName[0] == '.')//过滤.和..                continue;            //wcout<<pNextInfo.cFileName<<endl;            j++;            printf("%s\n",WcharToChar(pNextInfo.cFileName));            //对读入的图片进行处理            Mat srcImage = imread( getstring(i) + "/" + WcharToChar(pNextInfo.cFileName),CV_LOAD_IMAGE_GRAYSCALE);            Mat trainImage;            resize(srcImage,trainImage,Size(image_cols,image_rows),(0,0),(0,0),CV_INTER_AREA);//使用象素关系重采样。当图像缩小时候，该方法可以避免波纹出现           // threshold(trainImage,trainImage,0,255,CV_THRESH_BINARY|CV_THRESH_OTSU);            Canny(trainImage ,trainImage ,150,100,3,false);            for(int k = 0; k<image_rows*image_cols; ++k)            {                trainingData[i*sample_mun_perclass+(j-1)][k] = (float)trainImage.data[k];                //trainingData[i*sample_mun_perclass+(j-1)][k] = (float)trainImage.at<unsigned char>((int)k/8,(int)k%8);//(float)train_image.data[k];                //cout<<trainingData[i*sample_mun_perclass+(j-1)][k] <<" "<< (float)trainImage.at<unsigned char>(k/8,k%8)<<endl;            }        } while (FindNextFile(hFile,&pNextInfo) );    }    // Set up training data Mat    Mat trainingDataMat(class_mun*sample_mun_perclass, image_rows*image_cols, CV_32FC1, trainingData);    //cout<<"trainingDataMat:"<<endl;    //cout<<trainingDataMat<<endl;    cout<<"trainingDataMat——OK！"<<endl;    // Set up label data      for(int i=0;i<=class_mun-1;++i)    {        for(int j=0;j<=sample_mun_perclass-1;++j)        {            for(int k = 0;k<class_mun;++k)            {                if(k==i)                    labels[i*sample_mun_perclass + j][k] = 1;                else labels[i*sample_mun_perclass + j][k] = 0;            }        }    }    // Set up label data     Mat labelsMat(class_mun*sample_mun_perclass, class_mun, CV_32FC1,labels);    cout<<"labelsMat:"<<endl;    cout<<labelsMat<<endl;    cout<<"labelsMat——OK！"<<endl;    //训练代码    cout<<"training start...."<<endl;    CvANN_MLP bp;    // Set up BPNetwork's parameters    CvANN_MLP_TrainParams params;    params.train_method=CvANN_MLP_TrainParams::BACKPROP;    params.bp_dw_scale=0.001;    params.bp_moment_scale=0.1;    params.term_crit = cvTermCriteria(CV_TERMCRIT_ITER|CV_TERMCRIT_EPS,10000,0.0001);  //设置结束条件    //params.train_method=CvANN_MLP_TrainParams::RPROP;    //params.rp_dw0 = 0.1;    //params.rp_dw_plus = 1.2;    //params.rp_dw_minus = 0.5;    //params.rp_dw_min = FLT_EPSILON;    //params.rp_dw_max = 50.;    //Setup the BPNetwork    Mat layerSizes=(Mat_<int>(1,4) << image_rows*image_cols,int(image_rows*image_cols/2),int(image_rows*image_cols/2),class_mun);    bp.create(layerSizes,CvANN_MLP::SIGMOID_SYM,1.0,1.0);//CvANN_MLP::SIGMOID_SYM                                               //CvANN_MLP::GAUSSIAN                                               //CvANN_MLP::IDENTITY    cout<<"training...."<<endl;    bp.train(trainingDataMat, labelsMat, Mat(),Mat(), params);    bp.save("bpcharModel.xml"); //save classifier    cout<<"training finish...bpModel1.xml saved "<<endl;    //测试神经网络    cout<<"测试："<<endl;    Mat test_image = imread("25.jpg",CV_LOAD_IMAGE_GRAYSCALE);    Mat test_temp;    resize(test_image,test_temp,Size(image_cols,image_rows),(0,0),(0,0),CV_INTER_AREA);//使用象素关系重采样。当图像缩小时候，该方法可以避免波纹出现    //threshold(test_temp,test_temp,0,255,CV_THRESH_BINARY|CV_THRESH_OTSU);    Canny(test_temp ,test_temp ,150,100,3,false);    Mat_<float>sampleMat(1,image_rows*image_cols);     for(int i = 0; i<image_rows*image_cols; ++i)      {          sampleMat.at<float>(0,i) = (float)test_temp.data[i];       // sampleMat.at<float>(0,i) = (float)test_temp.at<uchar>(i/8,i%8);  //(float)resizeImage.data[k]    }      Mat responseMat;      bp.predict(sampleMat,responseMat);     float* p=responseMat.ptr<float>(0);      float max= -1,min =0;      int index = 0;      for(int k=0;k<class_mun;++k)      {          cout<<(float)(*(p+k))<<" ";          if(k==class_mun-1)              cout<<endl;          if((float)(*(p+k))>max)          {              min = max;              max = (float)(*(p+k));              index = k;          }          else          {              if(min < (float)(*(p+k)))                  min = (float)(*(p+k));          }      }      //对应上美、丑    string judge = "";    if (index==0)        judge = "美美的！";    if (index==1)        judge = "略丑啊！";    cout<<"识别结果:"<<judge<<endl<<"识别置信度:"<<(((max-min)*100) > 100 ? 100:((max-min)*100))<<endl;    /*Point maxLoc;    double maxVal = 0;    minMaxLoc(responseMat,NULL,&maxVal,NULL,&maxLoc);    cout<<"识别结果："<<maxLoc.x<<"  置信度:"<<maxVal*100<<"%"<<endl;*/    imshow("test_image",test_image);      imshow("test_temp",test_temp);      waitKey(0);    return 0;}//读取目录src下min（number，所有图像）图像提取人脸并保存到srccut目录，//参数：原图片目录src       剪切图片保存目录dst     读取最大数量number    void AllImagePro( string src, string dst, static int number ){    int count=0;    string src1 = src;      string src1cut = dst;    HANDLE hFile;    LPCTSTR lpFileName = StringToWchar(src1+"/"+"*.*"); //指定搜索目录和文件类型，如搜索d盘的音频文件可以是"D:\\*.mp3"    WIN32_FIND_DATA pNextInfo;  //搜索得到的文件信息将储存在pNextInfo中;    hFile = FindFirstFile(lpFileName,&pNextInfo);//请注意是 &pNextInfo , 不是 pNextInfo;    if(hFile == INVALID_HANDLE_VALUE)    {        //搜索失败        exit(-1);    }    cout<<"文件夹"<<src<<"找到的图片："<<endl;    do    {        if(pNextInfo.cFileName[0] == '.')//过滤.和..            continue;        count++;        printf("%s\n",WcharToChar(pNextInfo.cFileName));        Mat img = imread( src1 + "/" + WcharToChar(pNextInfo.cFileName) , 1 );        detectAndCut( img ,src1cut ,WcharToChar(pNextInfo.cFileName) );    }while (FindNextFile(hFile,&pNextInfo) && count<number);//如果设置读入的图片数量，则以设置的为准，如果图片不够，则读取文件夹下所有图片}//人脸检测//参数：待检测图像img       保存路径dir     保存文件名namevoid detectAndCut( Mat img ,string dir, string filename){   std::vector<Rect> faces;   Mat img_gray;   cvtColor( img, img_gray, COLOR_BGR2GRAY );   equalizeHist( img_gray, img_gray );   //-- Detect faces   face_cascade.detectMultiScale( img_gray, faces, 1.1, 2, 0|CV_HAAR_SCALE_IMAGE, Size(30, 30) );   for( size_t i = 0; i < faces.size(); i++ )    {      Point rec( faces[i].x, faces[i].y );      Point rec2( faces[i].x + faces[i].width, faces[i].y + faces[i].height );      Mat roi_img = img( Range(faces[i].y,faces[i].y + faces[i].height), Range(faces[i].x,faces[i].x + faces[i].width) );      imwrite( dir+"/"+filename, roi_img );      }}char* WcharToChar(const wchar_t* wp)  {      char *m_char;    int len= WideCharToMultiByte(CP_ACP,0,wp,wcslen(wp),NULL,0,NULL,NULL);      m_char=new char[len+1];      WideCharToMultiByte(CP_ACP,0,wp,wcslen(wp),m_char,len,NULL,NULL);      m_char[len]='\0';      return m_char;  }  wchar_t* CharToWchar(const char* c)  {       wchar_t *m_wchar;    int len = MultiByteToWideChar(CP_ACP,0,c,strlen(c),NULL,0);      m_wchar=new wchar_t[len+1];      MultiByteToWideChar(CP_ACP,0,c,strlen(c),m_wchar,len);      m_wchar[len]='\0';      return m_wchar;  }  wchar_t* StringToWchar(const string& s)  {      const char* p=s.c_str();      return CharToWchar(p);  }  string getstring ( const int n ){    std::stringstream newstr;    newstr<<n;    return newstr.str();}

美女原图：

NOT美女原图：

进行人脸检测截取后：
美女训练集：

NOT美女训练集：

照片都是百度随便搜的……

测试结果：

识别正确率80%左右
主要原因：
1、面部提取的不是特别准确，感觉略大
2、训练集太少，因为只是为了玩玩，只有16张图做训练集，所以会一定程度上影响测试的准确性
3、图像中脸部有的倾斜太严重，妹子们拍照总喜欢歪着脸，……会影响结果
4、所有人脸测试时均标准化为30*30，略小
5、采用canny算子得到的边缘进行训练，忽略了气色等因素，脸白脸黑其实边缘都差不多