how to get the time for feature extraction

classic Classic list List threaded Threaded
4 messages Options
Reply | Threaded
Open this post in threaded view
|

how to get the time for feature extraction

guan
Hello matlabbe,

I want to get the time for feature extraction to compare the performance between SIFT and SUFT.

I found two things: Keypoints 3D and Keypoints detection in  Statistics window, and I can get the figure about these, but there are no notes in Statistics Class. So please can you tell me I am right?  Thx!
Reply | Threaded
Open this post in threaded view
|

Re: how to get the time for feature extraction

matlabbe
Administrator
Hi,

If it is only to compare features extraction time from different algorithms, using find-object may be more easier to use.

In RTAB-Map, you should look under the Statistics view / TimingMem tab. Look for "Keypoints detection" and "Descriptors extraction". The number of features extracted is shown under "keypoint" tab, label "Keypoints count in the last signature".

cheers

Reply | Threaded
Open this post in threaded view
|

Re: how to get the time for feature extraction

guan
Thanks, matlabbe!

I use different feature type to test, and I think GFIT+BRIEF is a good choice.
Reply | Threaded
Open this post in threaded view
|

Re: how to get the time for feature extraction

matlabbe
Administrator
This post was updated on .
Well that depends for which purpose. For odometry, GFTT+BRIEF works well as we need something fast and the scale/rotation between two consecutive images don't change too much. SIFT and SURF are rotation/scale invariant, which I find better for loop closure detection over large environments.

cheers