apeker's profile - activity

Hi there I am trying to implement am image reverse search engine using non patented algorithms with JAVA.

I have been testing BRISK+FREAK, BRISK+BRIEF, BRISK+BRISK etc.

MY last tests were with BRISK+BRIEF which provided a good set of results (though I am not 100% sure about scaled then cropped images).

I have tried with 10.000 clusters (10.000 image features obtained by BRISK detector + BRIEF extractor) with BruteForce Hamming:

DescriptorMatcher.create(DescriptorMatcher.BRUTEFORCE_HAMMING)

and using

matcher.knnMatch(query,matches,1);

to find matches .

Though the results are quite good but I can not get it work with more than 10.000. Because If I try with more then 10.000 images I get the error of:

OpenCV Error: Assertion failed *((int64)imgCountIMGIDX_ONE < INT_MAX)**

While I write my question I came a cross of another question:

Running into an error using cv::BFMatcher (bruteforce matcher) with 10,100 training images

But there is no answer there either :(

I am trying everything from a 64bit windows machine. Any answer about this will be really helpful.