MalashinRoman's profile - activity

I have exactly the same problem. When I do bowtrainer.cluster() I get the runtime error on insufficient memory. It somehow deals with cv::Mat implementation because I get the error before cluster() if I collect all the descriptors in one Mat passing them via add() command.

I can collect and process no more than 500 000 descriptors (the process uses about 300 MB), If I collect more program fails:(