PHE's profile - activity

Hi all,

I am trying to detect similarities between images which come in an ongoing stream of images. Basically I want to know, if the system has already seen an incoming image before. The similarity comparison has to be at least rotation invariant. Therfore I am using ORB features, and perform a one to one comparison of the descriptors using FlannBasedMatcher. This works fine so far.

From what I have read on various sources on the web, I assume that I can also use Flann to make a one image vs. multiple images comparison. Right? I presume that I have to use somehow the flann.add and flann.train functions. Unfortunately I can not find any documentation on how to do that using Python. Any hints and help how this has to be done, or where to find further documentation on that is very much appreciated.

Thanks for your help.