This forum is disabled, please visit https://forum.opencv.org
 | 1 | initial version |
Which clustering method did you mean? Your question is a little bit vague.
flann can handle binary features well, but the centers which you'll get will be in float (the same would probably happen if you modify the OpenCV-kmeans).
Or do you actually just need the nearest neighbor for binary features? Then you can use the BruteForceMatcher-class in conjunction with Hamming-distance or flann with Hamming or LSH.
Afaik, sth like k-medoids (where the output-centers are discrete features) is currently not implemented in OpenCV, neither is agglomerative hierarchical clustering. I am also waiting that this will be part of OpenCV sometime...
