clustering ORB features
I am working on an application which uses Bag of Words model. I couldn't find a way to cluster binary features like ORB. BOWKMeansTrainer gives error if mat type is CV_8UC1. Can we cluster binary features like ORB? If yes, is there an efficient implementation available?
This questions is kinda related to http://answers.opencv.org/question/2601/can-the-bag-of-words-functions-be-used-with-binary/ . It doesn't seem that it is possible by opencv, the hierarchicalClustering method of flann computes centers, too, but they are not binary any more but float (unfortunately the function only accepts a float-Matrix for the centers) and it is unclear to me, if a simple cast to uchar will do it.