Speed up the estimateFu​ndamentalM​atrix step using parallelism

Hi! I am trying to create 3d scenes using structure from motion, but the function estimateFundamentalMatrix takes a lot of time to compute the fundamental matrix. The number of images that I need to perform this computation on is huge. Is there a way to parallelize the process or make it run on a GPU or something?

Please let me know if any kind of speed up is at all possible!!

Thanks in advance!!

[Qu Cao]

You can try tuning the parameters of the function to speed up, for example, increasing 'DistanceThreshold' and reducing 'Confidence'.

[Siddharth Shah]

Yes, in the end if nothing else works out, then as a last resort I will certainly do that. The only issue is that tuning the parameters to make the computation faster will decrease the accuracy of the function. And the nature of my project is such that compromising with accuracy might be detrimental.

But if that is the only option left, then yeah, I'll continue with it!

Thanks a lot!