1
$\begingroup$

What is the best possible runtime for extracting SIFT keypoints and Descriptors?
I know it depends on #keypoints extracted. So, say for image of size 640x640, the code that I have been using requires around 3 seconds for description of ~2000 keypoints.

What is the best runtime achieved and how (on similar grounds)? How can I make it run in realtime?
Is hadoop with 100 clusters a solution?

|improve this question
$\endgroup$
0
$\begingroup$

SIFT implemented on a GPU can run in realtime!

http://cs.unc.edu/~ccwu/siftgpu/

|improve this answer
$\endgroup$
  • 1
    $\begingroup$ What is the best runtime we can get for ~2000 keypoints on ~2.2GHz and RAM of 4GB? I have been using C++ code by Andrew Vedaldi and it takes around 3.5s for 2000 keypoints. Is there any implementation which does this within 1s with above configuration? $\endgroup$ – object recognition Jun 19 '14 at 6:22

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.