Different results each loop

asked 2019-03-17 06:48:01 -0500

droog_the_droog gravatar image

updated 2019-03-17 07:03:05 -0500

Hello,

I'm using KAZE for computing keypoints and descriptors for a set of images. If I just wrap my program in a while-loop, the program will generate different results for the same images each time. I calculate a matching score (according to Mikolajczyk et al), which becomes an easy way to see that results differ each iteration. This is super unintuitive for me, and I really have no idea what could be the problem. Is it like some issue with overwriting memory?

If anyone has any idea what could possibly cause an issue like this, I would be extremely grateful.

Best, Nik

PS: I'm using the Python bindings for scripting.

edit retag flag offensive close merge delete

Comments

I seem to be getting different matches every time I call knnMatch(descriptors1,descriptors2,k=2). Trying to figure out why...

droog_the_droog gravatar imagedroog_the_droog ( 2019-03-17 08:11:04 -0500 )edit

Hi Nik, there is any "random event" inside the scoring function? Probably the different result are simply because a sampling is used so every time different points are chosen. This is what I do for example in my algorithm. Btw I suggest you to switch to AKAZE, is fastest and better most of the time, you can get an overview of them here and here

HYPEREGO gravatar imageHYPEREGO ( 2019-03-18 05:46:49 -0500 )edit

Yeah, it's because of the randomness of the search algorithm I'm using, I've found. Also, KAZE does much better for my application in particular, but I really appreciate your answer regardless. :)

droog_the_droog gravatar imagedroog_the_droog ( 2019-03-18 09:31:39 -0500 )edit