CODEWITHSUNDEEP
opencvk-means
kdottiemo
kdottiemo

Reputation: 79

OpenCV kmeans: N>=K exception, error (-215)

when I try to use kmeans as such:

int K = 4;
Mat labels;
Mat centers;
std::vector<float> values;

// (put a bunch of values into "values" here...)

kmeans(values, K, labels, TermCriteria(TermCriteria::COUNT + TermCriteria::EPS, 10, 1.0), 10, KMEANS_PP_CENTERS, centers);

I get the error: "error: (-215) N >= K in function kmeans"

values.size() = 360000, so N is clearly greater than K. What gives? Thanks.

Upvotes: 0

Views: 4207

Answers (2)

Ashwini Pai
Ashwini Pai

Reputation: 23

I faced the same problem.

This worked for me. Converted vector to Mat.

int K = 4;
Mat labels;
Mat centers;
std::vector<float> values;
Mat points=Mat(values,true);
points.convertTo(points, CV_32F);
kmeans(points, K, labels, TermCriteria(TermCriteria::COUNT + TermCriteria::EPS, 10, 1.0), 10, KMEANS_PP_CENTERS, centers);

Upvotes: 0

Andrey Kamaev
Andrey Kamaev

Reputation: 30122

OpenCV weirdly interprets one-dimensional data as a 1 element array.

Something like following should fix the behavior:

kmeans(cv::Mat(values).reshape(1, values.size()), K, labels, TermCriteria(TermCriteria::COUNT + TermCriteria::EPS, 10, 1.0), 10, KMEANS_PP_CENTERS, centers);

Upvotes: 3

Related Questions