• Fisherfaces 算法的具体实现源码


      1 /*
      2  * Copyright (c) 2011. Philipp Wagner <bytefish[at]gmx[dot]de>.
      3  * Released to public domain under terms of the BSD Simplified license.
      4  *
      5  * Redistribution and use in source and binary forms, with or without
      6  * modification, are permitted provided that the following conditions are met:
      7  *   * Redistributions of source code must retain the above copyright
      8  *     notice, this list of conditions and the following disclaimer.
      9  *   * Redistributions in binary form must reproduce the above copyright
     10  *     notice, this list of conditions and the following disclaimer in the
     11  *     documentation and/or other materials provided with the distribution.
     12  *   * Neither the name of the organization nor the names of its contributors
     13  *     may be used to endorse or promote products derived from this software
     14  *     without specific prior written permission.
     15  *
     16  *   See <http://www.opensource.org/licenses/bsd-license>
     17  */
     18 
     19 #include "opencv2/core/core.hpp"
     20 #include "opencv2/contrib/contrib.hpp"
     21 #include "opencv2/highgui/highgui.hpp"
     22 
     23 #include <iostream>
     24 #include <fstream>
     25 #include <sstream>
     26 
     27 using namespace cv;
     28 using namespace std;
     29 
     30 static Mat norm_0_255(InputArray _src) {
     31     Mat src = _src.getMat();
     32     // Create and return normalized image:
     33     Mat dst;
     34     switch(src.channels()) {
     35     case 1:
     36         cv::normalize(_src, dst, 0, 255, NORM_MINMAX, CV_8UC1);
     37         break;
     38     case 3:
     39         cv::normalize(_src, dst, 0, 255, NORM_MINMAX, CV_8UC3);
     40         break;
     41     default:
     42         src.copyTo(dst);
     43         break;
     44     }
     45     return dst;
     46 }
     47 
     48 static void read_csv(const string& filename, vector<Mat>& images, vector<int>& labels, char separator = ';') {
     49     std::ifstream file(filename.c_str(), ifstream::in);
     50     if (!file) {
     51         string error_message = "No valid input file was given, please check the given filename.";
     52         CV_Error(CV_StsBadArg, error_message);
     53     }
     54     string line, path, classlabel;
     55     while (getline(file, line)) {
     56         stringstream liness(line);
     57         getline(liness, path, separator);
     58         getline(liness, classlabel);
     59         if(!path.empty() && !classlabel.empty()) {
     60             images.push_back(imread(path, 0));
     61             labels.push_back(atoi(classlabel.c_str()));
     62         }
     63     }
     64 }
     65 
     66 int main(int argc, const char *argv[]) {
     67     // Check for valid command line arguments, print usage
     68     // if no arguments were given.
     69     if (argc < 2) {
     70         cout << "usage: " << argv[0] << " <csv.ext> <output_folder> " << endl;
     71         exit(1);
     72     }
     73     string output_folder = ".";
     74     if (argc == 3) {
     75         output_folder = string(argv[2]);
     76     }
     77     // Get the path to your CSV.
     78     string fn_csv = string(argv[1]);
     79     // These vectors hold the images and corresponding labels.
     80     vector<Mat> images;
     81     vector<int> labels;
     82     // Read in the data. This can fail if no valid
     83     // input filename is given.
     84     try {
     85         read_csv(fn_csv, images, labels);
     86     } catch (cv::Exception& e) {
     87         cerr << "Error opening file "" << fn_csv << "". Reason: " << e.msg << endl;
     88         // nothing more we can do
     89         exit(1);
     90     }
     91     // Quit if there are not enough images for this demo.
     92     if(images.size() <= 1) {
     93         string error_message = "This demo needs at least 2 images to work. Please add more images to your data set!";
     94         CV_Error(CV_StsError, error_message);
     95     }
     96     // Get the height from the first image. We'll need this
     97     // later in code to reshape the images to their original
     98     // size:
     99     int height = images[0].rows;
    100     // The following lines simply get the last images from
    101     // your dataset and remove it from the vector. This is
    102     // done, so that the training data (which we learn the
    103     // cv::FaceRecognizer on) and the test data we test
    104     // the model with, do not overlap.
    105     Mat testSample = images[images.size() - 1];
    106     int testLabel = labels[labels.size() - 1];
    107     images.pop_back();
    108     labels.pop_back();
    109     // The following lines create an Fisherfaces model for
    110     // face recognition and train it with the images and
    111     // labels read from the given CSV file.
    112     // If you just want to keep 10 Fisherfaces, then call
    113     // the factory method like this:
    114     //
    115     //      cv::createFisherFaceRecognizer(10);
    116     //
    117     // However it is not useful to discard Fisherfaces! Please
    118     // always try to use _all_ available Fisherfaces for
    119     // classification.
    120     //
    121     // If you want to create a FaceRecognizer with a
    122     // confidence threshold (e.g. 123.0) and use _all_
    123     // Fisherfaces, then call it with:
    124     //
    125     //      cv::createFisherFaceRecognizer(0, 123.0);
    126     //
    127     Ptr<FaceRecognizer> model = createFisherFaceRecognizer();
    128     model->train(images, labels);
    129     // The following line predicts the label of a given
    130     // test image:
    131     int predictedLabel = model->predict(testSample);
    132     //
    133     // To get the confidence of a prediction call the model with:
    134     //
    135     //      int predictedLabel = -1;
    136     //      double confidence = 0.0;
    137     //      model->predict(testSample, predictedLabel, confidence);
    138     //
    139     string result_message = format("Predicted class = %d / Actual class = %d.", predictedLabel, testLabel);
    140     cout << result_message << endl;
    141     // Here is how to get the eigenvalues of this Eigenfaces model:
    142     Mat eigenvalues = model->getMat("eigenvalues");
    143     // And we can do the same to display the Eigenvectors (read Eigenfaces):
    144     Mat W = model->getMat("eigenvectors");
    145     // Get the sample mean from the training data
    146     Mat mean = model->getMat("mean");
    147     // Display or save:
    148     if(argc == 2) {
    149         imshow("mean", norm_0_255(mean.reshape(1, images[0].rows)));
    150     } else {
    151         imwrite(format("%s/mean.png", output_folder.c_str()), norm_0_255(mean.reshape(1, images[0].rows)));
    152     }
    153     // Display or save the first, at most 16 Fisherfaces:
    154     for (int i = 0; i < min(16, W.cols); i++) {
    155         string msg = format("Eigenvalue #%d = %.5f", i, eigenvalues.at<double>(i));
    156         cout << msg << endl;
    157         // get eigenvector #i
    158         Mat ev = W.col(i).clone();
    159         // Reshape to original size & normalize to [0...255] for imshow.
    160         Mat grayscale = norm_0_255(ev.reshape(1, height));
    161         // Show the image & apply a Bone colormap for better sensing.
    162         Mat cgrayscale;
    163         applyColorMap(grayscale, cgrayscale, COLORMAP_BONE);
    164         // Display or save:
    165         if(argc == 2) {
    166             imshow(format("fisherface_%d", i), cgrayscale);
    167         } else {
    168             imwrite(format("%s/fisherface_%d.png", output_folder.c_str(), i), norm_0_255(cgrayscale));
    169         }
    170     }
    171     // Display or save the image reconstruction at some predefined steps:
    172     for(int num_component = 0; num_component < min(16, W.cols); num_component++) {
    173         // Slice the Fisherface from the model:
    174         Mat ev = W.col(num_component);
    175         Mat projection = subspaceProject(ev, mean, images[0].reshape(1,1));
    176         Mat reconstruction = subspaceReconstruct(ev, mean, projection);
    177         // Normalize the result:
    178         reconstruction = norm_0_255(reconstruction.reshape(1, images[0].rows));
    179         // Display or save:
    180         if(argc == 2) {
    181             imshow(format("fisherface_reconstruction_%d", num_component), reconstruction);
    182         } else {
    183             imwrite(format("%s/fisherface_reconstruction_%d.png", output_folder.c_str(), num_component), reconstruction);
    184         }
    185     }
    186     // Display if we are not writing to an output folder:
    187     if(argc == 2) {
    188         waitKey(0);
    189     }
    190     return 0;
    191 }
  • 相关阅读:
    设计模式的四个基本要素
    拖拉记录上下移动--Ajax UI
    Rails-Treasure chest2 嵌套表单;
    YAML(摘录)
    Rails-Treasure chest1 (自定义Model网址;多语言包; 时区设置, TimeZone类; 格式日期时间; 表单单选UI; 表单多选UI;Select2 Plugin)
    iTerm2的设置和Zsh.
    **优化--后端**: 计数缓存counter_cache; rack-mini-profiler(2300🌟) ; bullet(5000✨):侦测N+1query
    优化--前端(全占课,未完成作业:);CDN; Http/2的设置(未完成)
    null值的判断
    if else
  • 原文地址:https://www.cnblogs.com/zzuyczhang/p/4457487.html
Copyright © 2020-2023  润新知