optimizing approach for SIFT detection

asked 2019-08-21 12:12:57 -0500

eric_engineer gravatar image

I'm running SIFT detection of a target in Python on Windows with a 2.7Ghz i7-6820HQ processor. I'm doing a sift.detectAndCompute, followed by FlannBasedMatcher, findHomograpy, and a perspective transform. It finds somewhere between 5-15 good matches.

It's processing 640x360px gray images in about 80-100ms. I'd like to improve that performance. Should I be looking at moving to a faster processor, switching to C++, going after some kind of GPU acceleration?

Thanks for any tips you can give.

edit retag flag offensive close merge delete