evalOCR.cpp

/*****************************************************************************
 *   Number Plate Recognition using SVM and Neural Networks
 ******************************************************************************
 *   by David Mill�n Escriv�, 5th Dec 2012
 *   http://blog.damiles.com
 ******************************************************************************
 *   Ch5 of the book "Mastering OpenCV with Practical Computer Vision Projects"
 *   Copyright Packt Publishing 2012.
 *   http://www.packtpub.com/cool-projects-with-opencv/book
 *****************************************************************************/

// Main entry code OpenCV

#include "OCR.h"

#include <opencv2/opencv.hpp>

#include <iostream>
#include <vector>

using namespace std;
using namespace cv;
OCR ocr;

void generateRandom(int n, int min, int max, vector<int>* samples)
{
    int range = max - min;
    int r = rand() % range + min;
    if (samples->at(r) == 0) {
        samples->at(r) = 1;
        n++;
    }
    if (n < 100) {
        generateRandom(n, min, max, samples);
    }
}

float test(Mat samples, Mat classes)
{
    float errors = 0;
    for (int i = 0; i < samples.rows; i++) {

        int result = ocr.classify(samples.row(i));

        if (result != classes.at<int>(i))
            errors++;
    }
    return errors / samples.rows;
}

int main(int argc, char** argv)
{

    int nh;
    char* data;
    // Check params
    if (argc >= 2) {
        nh = atoi(argv[1]);
        data = argv[2];
    } else {
        cout << "Usage:\n" << argv[0] << " <num hidden nodes> <data to load from xml>\n";
        return 0;
    }

    Mat classes;
    Mat trainingData;

    // Read file storage.
    FileStorage fs;
    fs.open("OCR.xml", FileStorage::READ);

    fs[data] >> trainingData;
    fs["classes"] >> classes;

    float result;

    // init random generator
    srand(time(NULL));
    // Create 100 random pos for samples
    std::vector<int> isSample(trainingData.rows, 0);
    generateRandom(0, 0, trainingData.rows - 1, &isSample);

    // Create sample data
    Mat train, trainClasses;
    Mat samples, samplesClasses;
    for (int i = 0; i < trainingData.rows; i++) {
        if (isSample[i] == 1) {
            samples.push_back(trainingData.row(i));
            samplesClasses.push_back(classes.row(i));
        } else {
            train.push_back(trainingData.row(i));
            trainClasses.push_back(classes.row(i));
        }
    }

    result = 0;

    ocr.train(train, trainClasses, nh);
    result = test(samples, samplesClasses);

    // cout << nh <<  "\t" << data << "\t"<< result <<"\n";
    cout << result;

    return 0;
}