C++ / OpenCV - Difference between Flann Index matching and Flann matching -
i trying undertstnad deeper level of feature matching using flann
, , looks there 2 usable approaches: , without index.
here so
question matching using flann
with indicies: how use opencv flann::index?
and here example of matching using flann
without indicies: https://github.com/itseez/opencv/blob/master/samples/cpp/matching_to_many_images.cpp
i see differences in code of course, i'm trying understand advantages of using 1 approach on other be. know in databases, adding index increases performance in many cases. analogous when using flann
match features???
does have experience this?
index feature matching way trade accuracy speed. flann called "approximate nearest neighbor search". means lose little bit accuracy (e.g. in fraction of times find next best nearest neighbor rather real best one), gain orders of magnitude speedup. because feature data noisy, approximation in searching stage tolerable. kgraph library provides index runs several times faster flann @ same accuracy.
Comments
Post a Comment