Fastest GPU implementation of a brute-force matcher for 128-float descriptors such as SIFT in 2NN mode, i.e., a match is returned if the best match between a query vector and a training vector is more than a certain threshold ratio better than the second-best match.

Float descriptors are slow. Check out my CUDAK2NN project for much faster binary description matching. Use a good binary descriptor such as LATCH where possible.

That said, this laboriously crafted kernel is EXTREMELY fast for a float matcher.

CUDA CC 3.0 or higher is required.

All functionality is contained in the files CUDAKfNN.h and CUDAKfNN.cu. 'main.cpp' is simply a sample test harness with example usage and performance testing.