Question

Im trying to implement the bag of words approach using opencv. After making the dictionary I am using the NormalBayesClassifier to train and predict the system.

I have prepared the trainme matrix as per the documentation as in each sample in each row. But the problem is that it gives an unhandled exception at this line: classifier.train(trainme, labels);

The complete code I am using is below:

int _tmain(int argc, _TCHAR* argv[])
{
initModule_nonfree();

Ptr<FeatureDetector> features = FeatureDetector::create("SIFT");
Ptr<DescriptorExtractor> descriptor = DescriptorExtractor::create("SIFT");
Ptr<DescriptorMatcher> matcher = DescriptorMatcher::create("FlannBased");

//defining terms for bowkmeans trainer
TermCriteria tc(CV_TERMCRIT_ITER+CV_TERMCRIT_EPS, 10, 0.001);
int dictionarySize = 100;
int retries = 1;
int flags = KMEANS_PP_CENTERS;
BOWKMeansTrainer bowTrainer(dictionarySize, tc, retries, flags);

BOWImgDescriptorExtractor bowDE(descriptor, matcher);

//**creating dictionary**//

Mat features1, features2;
Mat img = imread("c:\\1.jpg", 0);
Mat img2 = imread("c:\\2.jpg", 0);
vector<KeyPoint> keypoints, keypoints2;
features->detect(img, keypoints);
features->detect(img2,keypoints2);
descriptor->compute(img, keypoints, features1);
descriptor->compute(img2, keypoints2, features2);
bowTrainer.add(features1);
bowTrainer.add(features2);

Mat dictionary = bowTrainer.cluster();
bowDE.setVocabulary(dictionary);

//**dictionary made**//

//**now training the classifier**//

Mat trainme(0, dictionarySize, CV_32FC1); 
Mat labels(0, 1, CV_32FC1); //1d matrix with 32fc1 is requirement of normalbayesclassifier class

Mat bowDescriptor, bowDescriptor2;
bowDE.compute(img, keypoints, bowDescriptor);
trainme.push_back(bowDescriptor);
float label = 1.0;
labels.push_back(label);
bowDE.compute(img2, keypoints2, bowDescriptor2);
trainme.push_back(bowDescriptor2);
labels.push_back(label);

NormalBayesClassifier classifier;
classifier.train(trainme, labels);

//**classifier trained**//

//**now trying to predict using the same trained classifier, it should return 1.0**//

Mat tryme(0, dictionarySize, CV_32FC1);
Mat tryDescriptor;
Mat img3 = imread("2.jpg", 0);
vector<KeyPoint> keypoints3;
features->detect(img3, keypoints3);
bowDE.compute(img3, keypoints3, tryDescriptor);
tryme.push_back(tryDescriptor);

cout<<classifier.predict(tryme);
waitKey(0);



return 0;
}

Answer 1

I managed to figure it out, the problem lay here: float label = 1.0; as all the images being trained cannot have the same label. The system must be able to distinguish between the images given, thus its best to arrange the images in groups and give the groups the float values.